Changelog

12.11.0 (2026-08-28)

Features

• pr: scope gate-retry code review rounds to the fix-diff (#2089) (98409fa)

Bug Fixes

• scripts: attribute plugin_version per session instead of at extraction time (#2091) (5084ae6)

Code Refactoring

• agents: shrink complexity-review to its judgment residue (#2092) (a8be963)

12.10.0 (2026-08-27)

Features

• scripts: bridge recon artifact, cardinality tests, timeout pinning (#2070) (bd51330)

Bug Fixes

• scripts: Wave 1 of the open-issue consolidation plan (#2083) (efeda13)
• scripts: Wave 2 of the open-issue consolidation plan (#2084) (4428b5f)

Code Refactoring

• scripts: wave 4 of the open-issue consolidation plan (#2088) (4ff84e4)

Documentation

• adr: record no-go decision on splitting orchestrator.py (#2066) (885051a)

12.9.0 (2026-08-26)

Features

• context-ceiling: stop blocking agent dispatches, add the session-economy playbook (#2063) (5d0d58a)

Bug Fixes

• code-review: instrument review-agent JSON-contract parse failures (#2061) (c8f5f4d)

12.8.1 (2026-08-26)

Bug Fixes

• hooks: raise the absolute context ceiling from 150K to 350K (#2056) (a98bdab)

12.8.0 (2026-08-26)

Features

• metrics: report real per-agent context volume, not a message count (#2029) (2c79dc4)

Bug Fixes

• hooks: stop measuring subagent turns as main-thread context occupancy (#2054) (c9232bd)
• mutation-kill: four shard-pipeline and retry defects, plus yield steering (#2053) (b3958f2)

Performance Improvements

• agents: move orchestrator per-phase detail behind the knowledge index (#2034) (2456827), closes #2011
• mutation-kill: sequential concurrency default and target-aware stop policy (#2052) (4c266ce)

Documentation

• workflow: ratchet twice-reported mechanical findings into repo_invariants (#2026) (56e38af)

12.7.0 (2026-08-25)

Features

• code-review: economize the review panel (p1 slices) (#1987) (b722ede)
• scripts: add shippable session-report extractor for downstream users (#1984) (277311e)
• scripts: scope extract_session_report.py's output name, time range, and plugin version (#1986) (5b3a0c7)
• static-analysis: naming, accessibility, and performance probes (#1989) (4e87c19)

Bug Fixes

• code-review: make the shared context pack opt-in (#2025) (32204be)
• hooks: block by default at the context ceiling (#2022) (c51ebf7)
• hooks: use portable mktemp so the pre-push gate works on macOS (#1996) (25d68f2)
• metrics: add review-value sample validity check (#2020) (49b064c)
• scripts: count subagent transcripts in extract_session_report.py (#1995) (a3e4608)
• scripts: count subagent transcripts in session_extract.py (#2017) (ab786d3)
• scripts: simplify extract_session_report.py's --since to a day count (#1988) (cbf449a)

Performance Improvements

• code-review: build one shared context pack per review panel (#2021) (1895146)

12.6.0 (2026-08-24)

Features

• mutation-testing: extract mutation-report clustering and static-mutant filtering into Python (#1949) (ce09f52)
• test-improve: cut workflow token cost without weakening risk gates (#1966) (2bd0f1d)

Bug Fixes

• code-review: delete shared-classifier fallbacks, fail the import loudly (#1969) (8cc2d23), closes #1968
• mutation-kill: clean up exit-code and exception semantics (#1957) (1556958)
• mutation-kill: extend round-1 baseline-report reuse to --concurrency > 1 (#1942) (9f3da7a)
• mutation-kill: harden retry/downgrade mechanism (#1938) (#1954) (153c389)
• mutation-testing: fold static-skip survivors into the accepted/reason model (#1970) (b85247c)
• scripts: repair and extend the token-estimate drift gate (measure_tokens.py) (#1947) (60cd884)
• tests: make stryker shim scaffold-failure test privilege-independent (#1946) (6788ffd)

Code Refactoring

• mutation-kill: extract retry/downgrade module, merge Parallelism docs (#1934) (9bed5da)

12.5.0 (2026-08-06)

Features

• coverage-discovery: discover and classify Java multi-module test modules (#1838) (e9bafd1), closes #1765
• fix: add /fix workflow — triage-driven RED/GREEN defect fix, then PR (#1902) (40d64c2)
• mutation-kill: batch hardening from mutation-score-improvement analysis (#1931) (36f3ccd)

Bug Fixes

• batch: close 10 issues from issue-triage #1809-#1904 (#1916) (3b2fd10)
• batch: close 37 confirmed issues from issue-triage #1795-#1896 (#1911) (e0e4f36)
• cost-meter: add claude-opus-5 pricing and surface unpriced models durably (#1842) (3f0c467), closes #1830
• hooks: add bounded fail-open lock-acquisition budget to atomic_state.locked_state (#1894) (28ffb47), closes #1888
• hooks: lock boundary_events.py's log append against concurrent writers (#1891) (418c0c9), closes #1874
• hooks: lock JSONL appends against concurrent writers (#1898) (f73bf38), closes #1889
• scripts: skip opus-tier always-lenses on non-executable-only diffs (#1932) (0431e22), closes #1923
• skills: mutation-night-watch scheduling.md points at plugin root, not operator's repo (#1884) (b58d090), closes #1877

Code Refactoring

• test-improve: split SKILL.md into lean entry + per-phase references (#1921) (e865355)

Documentation

• orchestrator: document that parallel fix subagents must not dispatch their own review (#1882) (186cd5b)

12.4.0 (2026-08-05)

Features

• hooks: enforce Python-only shipped scripts in real time (#1769) (830ee68)
• mutation-testing: committed, human-approved mutation exclude-policy bootstrap (#1771) (09ce224)
• mutation-testing: stable dual-format mutation report with LATEST mirror (#1770) (ffee0e2)
• mutation-testing: unattended, LLM-free overnight mutation night-watch run (#1768) (e28ccd4)
• project-init: short-circuit re-invocation on already-initialized projects (#1821) (8eb203b), closes #1778

Bug Fixes

• auto-discover test/coverage projects instead of a frozen list (#1822) (4c5baf3)
• code-review: cap parallel dispatch at 10, retry once, never drop failures silently (#1760) (338c1fa), closes #1752
• code-review: mechanical backstops for dispatch-batching contract (#1761, #1762, #1763) (#1804) (93420aa)
• coverage-baseline: batch grp-1759 — .NET coverage discovery + 4 session-review skill fixes (#1810) (496905b)
• coverage-discovery: expand brace globs, normalize .sln separators, validate ** explicitly (#1833) (4bc67b3), closes #1827 #1828 #1832
• farley-score: clarify scoring criteria and add worked examples (#1816) (ff4315b), closes #1776
• hooks: dedupe redundant .review-passed read in cosmetic carry-forward lens (#1818) (45b1478), closes #1813
• hooks: detect equivalent git commit invocation forms in pre-commit gate (#1817) (a715451), closes #1801
• hooks: isolate test_emit_swallows_bad_cwd_type from real checkout (#1802) (d9188dc)
• hooks: remove second-boundary race in dispatch-window flaky test (#1811) (dd4befd)
• hooks: single-read-pass ledger query for cosmetic carry-forward veto (#1814) (6e85b9a), closes #1807
• session-review: guard the dev-checkout-only extractor dependency (#1820) (a73a0a5), closes #1779
• skills: enforce Conventional Commit titles in every issue-creating skill (#1792) (eee3cf1)
• stryker-xunit-v2-shim: gate Stryker on a recorded operator remediation choice (#1825) (d2b0a12)
• test-improve: target coverage work from per-module gap ranking (#1824) (c450f55)

Code Refactoring

• hooks: dedupe bypass-flag scan in pre_commit_detect (#1812) (ec97908)
• skills: dedupe severity vocab and raw-dir path in code-review consolidate (#1803) (3a20652)

12.3.0 (2026-08-03)

Features

• code-review: decouple drift-style review agents into /repo-review (#1742) (a8b5454)
• hooks: SessionStart nudge suggesting /repo-review on drift (#1740) (2d20b5a)

Bug Fixes

• hooks: pre_commit_review block messages should mention the --no-verify bypass (#1738) (3485e98)
• hooks: relocate repowise's tracked .mcp.json entry to local scope (#1749) (0f3e90e), closes #1747
• hooks: repo_review_nudge should trigger on % of codebase changed (#1744) (8531c05)

Documentation

• hooks: document the GATE_BYPASS_REASON pre-commit-review escape hatch (#1745) (df40d77)

12.2.0 (2026-08-02)

Features

• autoship: group related issues into shared spec/plan/PR batches (#1714) (f7e9409)
• orchestrator: make plan-quality drive code quality; publish cadence rule (#1728) (9cf8f52)
• scripts: wire orchestrator.py Implement phase to real persona dispatch (#1724) (ebb8a9c)
• scripts: wire orchestrator.py Plan phase to real persona dispatch (#1721) (08b64e1)
• scripts: wire orchestrator.py Research phase to real persona dispatch (#1718) (2329123)

Bug Fixes

• scripts: guard non-object dispatch envelopes in orchestrator.py (#1712) (e940c38)

Documentation

• extract shared review methodology to knowledge file (#1710) (4aa6c02)

12.1.0 (2026-08-02)

Features

• batch of 11 issue fixes across autoship, review gates, and content guards (#1704) (b9a4d86)

12.0.0 (2026-08-01)

⚠ BREAKING CHANGES

• agents: correctness-review's eval-corpus contract changes — cr-clean-baseline.json's expectedStatus flips from "pass" to "warn" now that category 6 can detect the two unprobed runtime claims already present in that fixture's target file. Per issue #101's eval-corpus-is-the-semver- contract convention, a flipped verdict in an existing fixture is a breaking change to the agent's observable behavior, not a patch.

Bug Fixes

• agents: add correctness-review's 6th/7th detect categories and protocol (#1697) (516cad0), closes #1639
• agents: settle category as the canonical finding taxonomy field (#1693) (2f877e0), closes #1639
• ci: raise the shipped Python floor from 3.8 to 3.10 (#1682) (6230ae8)
• hooks: harden the review-gate multi-line-literal normalizer (#1677) (7f3378d)
• scripts: revert Repowise's unexpected .claude/CLAUDE.md writes (#1688) (f224d91), closes #1670
• skills: Repowise's CLAUDE.md write is intended, not a bug (#1691) (be22cc8), closes #1670
• skills: resolve shipped-script paths and portability, batch 2 (#1683) (f88cdb5)
• tests: widen bare-invocation guard and quote CLAUDE_PLUGIN_ROOT expansions (#1687) (e2e3879)

Miscellaneous

• scripts: clean up recon and token-efficiency script nits (#1680) (50eb3c7), closes #1651

11.3.1 (2026-08-01)

Features

• evals: fix 8 more evals/expected/*.json fixture pointers still saying .js for .ts fixtures (#1641) (03742a2)

Bug Fixes

• agents: ship four agents' Implemented-by scripts inside the plugin (#1647) (441a9ec), closes #1636
• hooks: fix stacked heredoc opener and closer whitespace bugs in normalizer (#1659) (8551a79), closes #1638
• skills: /setup's mutmut install and version probes cannot succeed (#1673) (f7afbd7)
• skills: pin the scheduled-wake-up calling contract for long-running jobs (#1643) (42d2223)
• skills: six shipped skills invoke scripts by bare relative paths (#1658) (5a20f7a), closes #1637

11.3.0 (2026-07-31)

Features

• evals: fix js-fp-review eval-corpus hygiene gaps and repair issue-1620 fallout (#1632) (35df9b4)
• harness: reduce review-agent re-review churn — all six #1623 slices (#1631) (cbee175)

Bug Fixes

• tests: stop change_impact's real-diff test asserting on the current branch (#1640) (c807435)

11.2.0 (2026-07-31)

Features

• agents: extend js-fp-review with Mostly Adequate Guide concepts (#1619) (bcfe35a)

Bug Fixes

• mutation-kill: resolve six review findings from PR #1600 (#1615) (0facee8)

Code Refactoring

• dev-team: deduplicate knowledge across agents and skills (#1617) (dd750da)

11.1.1 (2026-07-31)

Bug Fixes

• code-review: review agents confabulate a missing diff instead of reporting the gap (#1595) (bdcc178), closes #1591
• mutation-kill: scope git commit/revert integrity, sanitize sln path, harden survivor-kill loop (#1599) (5343af1)

Code Refactoring

• mutation-kill: extract shared subprocess-timeout-and-log helper (#1597) (96c1b33), closes #1593

Miscellaneous

• code-review: deterministic pre-pass, fix-loop triage, de-dup mutation-kill git tests (#1612) (c9907f9)
• orchestrator: warn that disjoint final files don't excuse worktree isolation (#1613) (643aad2), closes #1609

11.1.0 (2026-07-30)

Features

• mutation-kill: reuse an available baseline mutation report for Round 1 (#1565) (f88a9e9)

Bug Fixes

• mutation-kill: --headless mode auto-commits LLM-generated test code with no content vetting (#1589) (54197e0)
• mutation-kill: add missing subprocess timeouts, fix xdist races, clamp budget_seconds (#1592) (cf5d1a7), closes #1572 #1574 #1578 #1549
• mutation-kill: ask the operator before degrading on time budget alone (#1552) (cc6cd27)

Code Refactoring

• mutation-kill: extract round-runner + split loop module (#1586) (40bf112)
• mutation-report: split parse_mutmut_junitxml's per-testcase status/dict-assembly logic (#1587) (4c914aa)

Performance Improvements

• mutation-kill: add timeouts to long-running subprocess.run calls (#1576) (7854dce)

Documentation

• document TaskCreate/ScheduleWakeup tool contracts + parallel-build git-race guidance (#1594) (35277e9), closes #1546 #1548 #1571

Miscellaneous

• mutation-feasibility-gate: type Decision.outcome as an Enum and group probe params (#1575) (29e6113)
• mutation-kill: consolidate identical dotnet_build_targets/dotnet_test_targets (#1581) (a4bcf09)

11.0.0 (2026-07-29)

⚠ BREAKING CHANGES

• the svelte-review review agent is removed from the plugin. Projects relying on it should use a11y-review + component-architecture-review for Svelte files.

Features

• build: default to sequential, make wide fan-out opt-in (#1517) (7da3139)
• build: gate inline review dispatch on diff domain via shared resolver (#1527) (1bf2647)
• code-review: adopt the shared select_lenses resolver in Step 3 (#1533) (379df81)
• cost-meter: persist per-subagent cache tokens in the durable log (#1532) (3c56ceb)
• cost-report: measure context pollution separately from one-time token spend (#1541) (681fa6f), closes #1520
• remove the svelte-review agent from the plugin (#1534) (eb46568)

Bug Fixes

• agents: scope git staging/commit commands to touched files outside worktree isolation (#1538) (38beaa8), closes #1519
• gherkin-derive: require endpoint-specific scenario titles (#1528) (e81955b), closes #1526
• gherkin-scripts: sanitize gherkin_stub_merge.py, widen bidi hardening (#1530) (5035e21), closes #1529

Documentation

• orchestrator: state explicit rule against full-transcript subagent status checks (#1539) (43a0feb), closes #1518
• review: add wave/review-agent fan-out consolidation guidance (#1540) (391fb24), closes #1521

10.25.1 (2026-07-28)

Bug Fixes

• hooks: serialize hook state read-modify-write to close lost-update races (#1507) (2c1859e), closes #1501
• test-improve: handle a malformed test-counts-before.json as absent in Phase 1 (#1508) (336f770), closes #1502

10.25.0 (2026-07-28)

Features

• coverage-baseline: guard and persist baseline at the tracked path (a5f2885), closes #1491
• coverage-delta: append coverage-history.json under tracked data/ (0aa2075)
• coverage-delta: persist mutation-history.json under tracked data/ (ff2799c)
• coverage-delta: read baseline-coverage.json from tracked data/ (27ffd88)
• quality-targets-converge: read coverage-history.json from tracked data/ (89a910e)
• quality-targets-converge: read mutation-history.json from tracked data/, atomic write-back (d69da67)
• test-improve: capture Phase 2 baselines directly to tracked data/ (8e8b193)

Bug Fixes

• hooks: agent_dispatch_ledger never matches plugin-qualified subagent_type (c9c1f99)
• lint: clear repo-wide ruff debt so the pre-push gate is green (d3e34a3)
• quality-targets-converge: point gherkin-effectiveness rollup at tracked baselines (90efe5f)
• quality-targets-converge: remove stale knob-7 opt-in language (7bda2fd)
• test-improve: correct Phase 8's whole-repo-splice source description (a685289)
• test-improve: correct two code-review findings from the final panel (ba89d06)
• test-improve: make the coverage-baseline cross-reference file-relative (5236f28)
• test-improve: point Phase 5 coverage-history mention at tracked data/ (318519b)
• test-improve: remove Phase 9's now-redundant copy-report-data step (24876a2)
• tests: resolve xdist parallel-worker races on shared /tmp and careful-state.json (9887aee)

Documentation

• correct report-output-location.md's #1412 rationale wording (eafe06b)

10.24.1 (2026-07-28)

Bug Fixes

• project-init: correct graphify/repowise setup docs and add missing gitignore rule (2e6fa25), closes #1483 #1484 #1486 #1487

10.24.0 (2026-07-27)

Features

• cd-test-architecture: wire and test the Step 1 surface-coverage gate (b6576db), closes #1464
• hooks: add fail-closed dispatch-ledger corroboration reader on metrics_query (4fc90f9)
• hooks: ledger genuine Agent-tool dispatches via a shared review-agent registry (6741c30)

Bug Fixes

• hooks: close dispatch-ledger forgery and subject-binding gaps found by security review (b7f6379)
• hooks: close remaining gaps found by four more security review rounds (0509271)
• hooks: require corroborating dispatch-ledger evidence for the review-passed gate (524b578)
• hooks: route git commit -a / pathspec commits through the dispatch-ledger gate (dca71c3), closes #1476
• hooks: stop referencing the retired agent_model_resolve.py hook (48101c0), closes #1475
• session-review: version-tag telemetry and default to local-only analysis (2d6bb63), closes #1471 #1480

Code Refactoring

• hooks: extract shared doc-classification constants (#1477) (fe1699d)
• hooks: extract shared git safety flags (#1477) (5a73567)
• hooks: split pre_commit_review.py's decision pipeline (#1477) (c9e75bc)
• scripts: parameterize gherkin_analysis_coverage_gate by heading level/section/categories (97f8077)

10.23.0 (2026-07-27)

Features

• gherkin-derive: dispatch two independent Gherkin quality review agents (1b79dcf)
• gherkin-public: dispatch two independent Gherkin quality review agents (344ba73), closes #1452
• gherkin-quality-review: add adversarial gap/balance review agent (0639058)
• gherkin-quality-review: add runtime isolation check and calibration provenance (f6ccee3)
• pr: add a PR template and make the shipped /pr skill template-aware (ce9037a), closes #1472
• rather than fix: per the repo's eval-corpus-semver contract. (6866b43)
• test-smell-review: flag mocking of internal-to-component collaborators (6866b43), closes #1451

Bug Fixes

• gherkin-derive: make comprehensive codebase-wide analysis the default (3a2710c)
• gherkin-quality-review: address real code-review findings before first PR (3973f16)
• gherkin-quality-review: stop pinning a model snapshot in the isolation-check script (e56f1b7)
• hooks: rewrite SessionStart hooks as Python; fix repo-wide ruff findings (b2c2544)
• lint: remove redundant quoted type annotations (UP037) (2a2d6ec), closes #1458
• software-engineer: surface test-double misuse guidance at authoring time (5814df4)
• test-doubles: address real code-review findings on internal-collaborator misuse (23a86fd)

Documentation

• knowledge: name internal-collaborator mocking as its own test-double misuse (f9eb231)

10.22.0 (2026-07-27)

Features

• apply-test-doubles: add report-vs-target resolution and cd-test-architecture fallback (94ab53c)
• apply-test-doubles: apply Step 4b's decision logic by citation, with --component scoping (9f07bb1)
• apply-test-doubles: wire dispatch mechanics and register the new skill (f2915c6)

Bug Fixes

• apply-test-doubles: clear stale Double-column tool name on revert (198f717)
• apply-test-doubles: correct dispatch mechanics after 2nd review round (7430fbf)
• apply-test-doubles: correct dispatch mechanics and 8 other review findings (c31e06b)

10.21.0 (2026-07-26)

Features

• cd-test-architecture: add downstream-service branch to Step 4b (532de15)
• cd-test-architecture: add downstream-service branch to Step 4b (ownership-independent) (8f3424c)
• cd-test-architecture: add downstream-service Story shapes (2a62193)
• cd-test-architecture: add downstream-service Story shapes (fa2fbf2)
• cd-test-architecture: add Step 4b build-vs-document scaffold (6bf14aa)
• cd-test-architecture: add virtual-service-libraries knowledge file (a4a81be)
• cd-test-architecture: branch Step 4b on testcontainers accept/decline for DB doubles (45151e2)
• cd-test-architecture: prefer virtual-service libraries over hand-rolled Fakes (5d01bba)
• cd-test-architecture: record downstream-service ownership in Step 1's inventory (eea24f0)
• cd-test-architecture: show build/document status in the report template (5fcce50)
• cd-test-architecture: write a test-double setup guide alongside the assessment report (c2d4be9)

Bug Fixes

• cd-test-architecture: address code-review findings on downstream-service branch (iteration 1) (ff867bb)
• cd-test-architecture: address code-review findings on redesign (c29e62c)
• cd-test-architecture: address code-review findings on test-double setup guide (fa0561e)
• cd-test-architecture: address code-review findings on virtual-service-library preference (3765432)
• cd-test-architecture: address review-fix iteration-2 findings on downstream-service branch (be9e946)
• cd-test-architecture: close code-review findings on Step 4b (f462f56)
• cd-test-architecture: close residual knowledge-file gaps (8bb5148)
• cd-test-architecture: close round-4 correctness cleanups on Step 4b (f5e7463)
• cd-test-architecture: declare --yes, fix stale comment (round-5 cleanup) (37aa6a5)
• cd-test-architecture: propagate virtual-service-library preference to remaining files (03f48ec)
• cd-test-architecture: reconcile knowledge-file placement guidance (e32798d)
• cd-test-architecture: resolve batching-shape and ambiguity-scope gaps (9f65760)
• cd-test-architecture: resolve chat-only path gap and citation drift in setup guide (483590d)
• knowledge: correct stale pre-merge guidance to prefer virtual-service libraries (4573c14)
• lint: fix pre-existing ruff findings and rename N999 probe modules (c37c995), closes #1438
• review-gate: move .review-passed to .claude/memory/ for centralized artifacts (40f7dfb)
• skills: migrate remaining skills off the legacy reports/ path (427e959), closes #1432

Code Refactoring

• cd-test-architecture: collapse Step 4b to one three-way per-component ask (e375f8a)

Miscellaneous

• cd-test-architecture: wip checkpoint for Step 4b scaffold (Step 1.1) (e7a307d)
• knowledge: reconcile remaining ownership-tiered placement wording (cd276d1), closes #1444

10.20.0 (2026-07-25)

Features

• gherkin-derive: detect and report possibly-stale retained scenarios (4bd10be)
• gherkin-derive: headline bdd-runner completion state, reconcile with Phase 5 (9d955af)
• gherkin-derive: read-before-write, merge existing .feature files (2a489fb)
• gherkin-derive: wire failure-path gate into Step 6 and the Phase-3 human gate (56ec6c2)
• hooks: add artifact_paths category directory accessors (52ca35a)
• hooks: add artifact_paths category_dir()/migrate_dir() for directory-level migration (3eaec1e)
• hooks: add artifact_paths.dev_team_reports_dir() accessor (4d5ea20)
• hooks: add artifact_paths.project_root() git-root resolver (7b1511f)
• hooks: add artifact_paths.resolve_file() per-file migration (beaa4af)
• hooks: add home-scoped telemetry consent helper (bbec3ad)
• hooks: one-time notice for inert legacy telemetry signal (a5e1543)
• scripts: add gherkin_failure_path_gate coverage gate (9f13654)
• scripts: add gherkin_feature_merge module (parser, merge, CLI, stale-check) (1a4b906)
• test-improve: track report data, fix stub merge resets, reorder phases for Gherkin signal (1dd421e)

Bug Fixes

• build: bookkeeping allowlist covers .claude/-nested paths (a7ea767)
• close code-review gaps in epic #1406's final documentation sweep (23625d5)
• close gaps found verifying the #1406 fix pass itself (2bb310a)
• code-review: sliced-mode ledger/consolidate relocate to .dev-team-reports/ (fb0c584)
• docs: dual-read fallback for cost-report/harness-audit's migrated metrics (f900f6e)
• docs: home-scope artifact-usage.json references to ~/.claude/metrics/ (8947168)
• gherkin-derive: file-relative cross-references, regenerate skills index (fa48fc2)
• hooks: close symlink, fail-open, and path-traversal gaps in artifact migration (557fa7e)
• hooks: default-off task/cost/session-learning writers, gated by telemetry consent (cc00c6f), closes #1407
• hooks: harden telemetry.py's marker file and log write-failure diagnostics (7e91eb7)
• hooks: migrate remaining Stop/SubagentStop and shared-lib writers to shared helper (098b98e)
• hooks: pre_commit_review gate-bypass-audit uses real project root (983335f)
• hooks: refactor-freeze guards use git-root resolution (6021b45)
• hooks: sweep production-code default paths to .claude/-nested locations (03cc2f8)
• hooks: telemetry consent is config-file-only, home-scoped (6e55811)
• project-init: gitignore .mcp.json when Repowise is provisioned (abca725), closes #1416
• relocate .gitignore rules for .claude/-nested paths and close AC11 gap (06354e3)
• scripts: address issues surfaced by fix-loop re-verification (57e17af)
• scripts: extract shared _gherkin_text helper, fix error-message and naming findings (8978d43)
• scripts: lazy-resolve gherkin rollup default path; empty stale allowlist (c398650)
• scripts: make gherkin_feature_merge's write atomic (ccbb53a)
• scripts: resolve code-review findings in gherkin-derive tooling (b7dd323)
• scripts: stop gherkin_feature_merge from corrupting files without a trailing newline (913050d)
• skills: migrate stale memory/ path refs, add missing skipped_duplicate_titles doc (8eaedda)

Code Refactoring

• scripts: extract shared vendored-tree pruning helper (160beed)
• scripts: move _vendored_tree.py to scripts/lib/, align its callers (f301f91)
• scripts: simplify redundant line-number arithmetic, fix lint findings (c934e67)

Documentation

• add operator-facing artifact-migration guidance (213ce2a)
• build: build-phase.json write instruction targets .claude/memory/ (d7f645f)
• build: gate review-value.jsonl on telemetry consent, relocate to .claude/metrics/ (abc8169), closes #1408
• code-review: interactive write path targets .dev-team-reports/ (3dcb6b1)
• dev-team: fix stale bare metrics/ and reports/ path references (a20b31a)
• report-output-location describes consolidated .dev-team-reports/ (ea8c28d)
• review-agent: write path targets .dev-team-reports/ (c5c6c07)
• scripts: document the merge command's TOCTOU limitation as a human decision (1654c7c)
• setup: Step 11 targets .claude/ and .dev-team-reports/, corrects stale env-var claims (f58f125)
• sweep remaining DEV_TEAM_REPORTS/ references to .dev-team-reports/ (9d13848)
• sweep remaining legacy artifact-path references repo-wide (Slice 7) (1936759)
• telemetry: document home-scoped consent-only on/off/status (47b0485)
• test-improve: describe gherkin-derive merge-and-report behavior (0e84025)
• test-improve: write-path instructions target .claude/ and .dev-team-reports/ domains (9d15aa9)
• triage: write path targets .dev-team-reports/ (6c5a32d)

Miscellaneous

• knowledge: rebuild index after gherkin-derive/test-improve SKILL.md edits (beb13b7)

10.19.1 (2026-07-24)

Bug Fixes

• unify MCP grant-check JSON schema and consolidate agent-audit docs (261b760)

Code Refactoring

• test-improve: simplify phase numbering to a flat sequence (9489479)

10.19.0 (2026-07-24)

Features

• add bdd-runner completion gate and security-assessment MCP tool checks (155bb0e)

Bug Fixes

• address code-review findings for gherkin gate and dropped implementer behavior (a9e6cb5)
• extract shared single-pass runner for MCP tool-grant check scripts (966f79a), closes #1392 #1393
• reconcile security-assessment MCP tool roster with the roster main already applied (75c3a13)
• resolve 492 pre-existing ruff findings across the repo's Python code (f71cf7d)
• session-review: correct two wrong step-number back-references (2b→3b, 0→1) and name Step 3 (0f5ac36)
• setup: gitignore .review-passed and .mcp.json by default (e7a018f)

Documentation

• cd-test-architecture: name Step 2b/Step 1 in back-references (27c9159)
• docker-image-audit: name Step 2b in the hadolint-limitations back-reference (f8c54b0)
• test-evaluation: name Step 2b and the cross-skill Step 3 in back-references (f151980)

Miscellaneous

• lint fix (7712986)

10.18.2 (2026-07-23)

Bug Fixes

• coverage-delta,quality-targets-converge: filter accepted survivors (d9aa92d)
• hooks: accumulate multi-tool usage into the turn-mark sentinel's tools_used list (13eb4a0)
• hooks: close sentinel duplication, coercion bug, and read-modify-write race from code review (94d779f)
• hooks: compose single- and multi-tool nudge messages with precedence-ordered cues (671ee5c)
• hooks: read multi-family turn-mark sentinel; register Repowise PostToolUse matcher (bc52091)
• hooks: route code-intelligence nudge across CodeGraph, Repowise, and Graphify (843c43c)
• mutation-kill: document accepted-survivor raw/adjusted score (f0bbdc1)
• mutation-testing: add accepted-survivor status, reason, dual score (8a8d7e7)

Code Refactoring

• hooks: extract shared turn-identity lib, rename codegraph_nudge to code_intelligence_nudge (4088d0d)
• hooks: rename codegraph_turn_mark to code_intelligence_turn_mark, adopt turn-identity lib (6b7d9c9)

Documentation

• hooks: fix cross-references and stale-ref test fixtures after the nudge-hook rename (263e4ab)
• hooks: rewrite the code-intelligence-nudge mechanism doc for three tools (308e61b)
• knowledge: refresh CodeGraph/Repowise tool surfaces, add routing precedence guidance (516621a)

Miscellaneous

• hooks: update registrations and cross-references for the renamed nudge hook (305ab47)

10.18.1 (2026-07-23)

Bug Fixes

• mutation-kill: protect the test file mutmut may corrupt; detect the Python 3.13+ pickle crash (83d484c)
• mutation-kill: restore os import dropped by main's ruff-wiring rebase (bb30b21)
• mutation-kill: stop test_mutation_gate.py from leaking ADAPTER_* env into later tests (ccfeaba)
• relocate graphify's machine-specific settings.json hook path (ffe5afe), closes #1367

10.18.0 (2026-07-23)

Features

• ci: wire ruff check into ci-local.sh's pre-push gate (e2ecbf4), closes #1362

10.17.0 (2026-07-22)

Features

• mutation-kill: extend the scripted mutant-kill loop to Python/mutmut (8b12208)
• setup: wire python/mutmut into Step 6 mutation-tool installer (763db3a)

Bug Fixes

• mutation-kill: revert the source file mutmut may leave mutated on crash (08027a5)
• mutation: mutmut adapter never detects survivors (deaec42)

10.16.0 (2026-07-22)

Features

• add agent-contract validator for frontmatter compliance (0224d27)
• add CodeGraph/Repowise guidance to apply-fixes skill (52ebbb1)
• add CodeGraph/Repowise guidance to build skill (fd3156d)
• add CodeGraph/Repowise guidance to cd-test-architecture skill (816eabf)
• add CodeGraph/Repowise guidance to ci-debugging skill (d40af80)
• add CodeGraph/Repowise guidance to domain-analysis skill (540e922)
• add CodeGraph/Repowise guidance to farley-score skill (c508f64)
• add CodeGraph/Repowise guidance to feature-file-validation skill (12f2075)
• add CodeGraph/Repowise guidance to gherkin-public skill (afda86e)
• add CodeGraph/Repowise guidance to hexagonal-architecture skill (2582717)
• add CodeGraph/Repowise guidance to legacy-code skill (49ba470)
• add CodeGraph/Repowise guidance to mutation-testing skill (ab4dc41)
• add CodeGraph/Repowise guidance to plan skill (1e39bd7)
• add CodeGraph/Repowise guidance to quality-targets-converge skill (0118479)
• add CodeGraph/Repowise guidance to semantic-duplication-scan skill (271da25)
• add CodeGraph/Repowise guidance to semantic-scan skill (ce2c38d)
• add CodeGraph/Repowise guidance to systematic-debugging skill (3f59646)
• add CodeGraph/Repowise guidance to test-audit-disable skill (8db5ef7)
• add CodeGraph/Repowise guidance to test-design-advisor skill (f3d3a2e)
• add CodeGraph/Repowise guidance to test-health skill (f200cc3)
• add CodeGraph/Repowise guidance to ubiquitous-language skill (14168e9)
• add diff-size fast path to /code-review agent eligibility (#1339) (cbcf779)
• discover async/event/scheduled surfaces in gherkin-derive (2b89255)
• don't treat existing tests as ground truth in gherkin-derive (c738611)
• ground gherkin-derive failure scenarios in real code branches (19a6ad3)
• measure effectiveness of gherkin-derive scenarios (282bc0a)
• migrate agent frontmatter to native model: and effort: high (e8fa54b)
• prefer CodeGraph/Repowise for gherkin-derive discovery and grounding (d23d524)
• require and assign color: on every agent per the fleet rule (#1334) (677c805)
• require skills: preload and memory: project on qualifying agents (#1335) (e256120)
• wire agent-contract validator into agent-audit, drop agent-eval calibration mode (6ef5f6e)

Bug Fixes

• address /build Step 6 backstop review findings on the fleet-conventions PR (73c45df)
• make doc-review defer to agent-contract.json over stale checker constants (ed36077), closes #1340
• repair dead links to the retired model-routing docs (57fd823)
• require fresh, correctly-scoped citations in the shared review-agent evidence check (9bee9bc), closes #1342
• reword /build's dangling /verify references to describe the pattern (311f623), closes #1341
• reword plan-review-acceptance TDD Step Traceability to Code-First terms (1654c8e), closes #1337

Code Refactoring

• convert implementer/spec-reviewer/quality-reviewer prompt templates to registered agents (4056fb5)
• convert plan-review personas from prompt templates to registered agents (9096cfe)
• wildcard the codegraph MCP tool grant in agent frontmatter (#1325) (f381dca)

Documentation

• generate the MkDocs nav from the docs tree via awesome-pages (70319b3), closes #1280

Miscellaneous

• delete orphaned plan-reviewer coordinator persona (#1331) (21befa1)
• move non-contract keys (cites, scope, enforcement) out of agent frontmatter into the body (dc9ea06), closes #1333
• retire band-to-model resolver hook, ladder, and calibration infrastructure (72c965a)

10.15.1 (2026-07-21)

Documentation

• supersede ADR 0024 to allow versioned model IDs; relax routing test (c3f0b4c)

Miscellaneous

• drop dead legacy tier alias keys from model-routing.json (3c18bc7), closes #1268
• pin latest model versions in model-routing.json (10d1f3a)

10.15.0 (2026-07-20)

Features

• add unattended /setup --yes flag that auto-confirms install gates (d1abe0a), closes #1234
• gate low-yield review lenses by change shape in /code-review (3653a64), closes #1254
• record severity and provenance in review-value.jsonl (c1f71c2), closes #1256 #1257

Bug Fixes

• /setup only installs tools that are missing (afbc5d1), closes #1236
• suppress empty SubagentStop heartbeats in task metrics (72abd97), closes #1258
• tolerate review-agent schema drift in /code-review aggregator (b43f0b4), closes #1261

Documentation

• correct dev-team counts, split /test-improve reference, fill workflow gaps (aae300a)
• fix dev-team-process doc drift and split band calibration (7b63352)
• fix eval/session/codegraph/test/telemetry doc drift (ffb765b)
• lead dev-team plugin install page with "install, then /setup" (900246a), closes #1231

10.14.2 (2026-07-20)

Bug Fixes

• correct invalid hook schema that broke plugin load (1b4cdd0), closes #1227

10.14.1 (2026-07-20)

Bug Fixes

• stop /project-init gating keyless Graphify build behind an API key (e7d9bc1), closes #1224

10.14.0 (2026-07-20)

Features

• add eval fixtures and floors coverage for uncalibratable agents; fix flapping fixtures (56db482)
• scope test-improve Phase 6 mutation validation to cumulative branch diff (26040c9)
• skills: add long-eval restart-durable eval skill (c87dce4)

Bug Fixes

• bump routing map to Claude 5 family, add 5-family pricing, align staleness doc (8591414)
• downgrade a11y-review to effort:low (calibration-verified) (c04620d)
• downgrade component-architecture-review to effort:low (calibration-verified) (ac14d08)
• downgrade concurrency-review to effort:low (PROVISIONAL — pending #1211) (94efe6d)
• downgrade correctness-review to effort:medium (calibration-verified) (06050bb)
• downgrade doc-review to effort:low (calibration-verified) (b256248)
• downgrade performance-review to effort:low (calibration-verified) (36f4549)
• downgrade refactor-opportunity-review to effort:low (calibration-verified) (bd02f0a)
• downgrade spec-compliance-review to effort:low (calibration-verified) (31902b7)
• revert correctness-review to effort:high (#1185 re-confirmation failed) (26a5b51)
• snapshot results list before aliasing it into the checkpoint dict (61eb9a2)

Documentation

• decouple calibration floor (consequence) from effort band (8062f90)

10.13.0 (2026-07-18)

Features

• add /run-report command joining boundary/cost/workflow-state streams (#1167) (a5e4ef2)
• add typed/faceted query helper over metrics streams (#1169) (f493942)
• hard per-iteration journal gate for autoship/ship loops (#1168) (9fa2d70)
• workflow state-machine event stream + dwell-time derivation (#1166) (9215b2e)

Bug Fixes

• add --json machine-readable mode to /review-agent for reliable cross-model output (c6aef13)
• record rubric effort-band decisions for progress-guardian and ui-ux-designer (a9f94f5)

Documentation

• note Dependabot-PR secret-access caveat for telemetry CI gate (66d5426)

Miscellaneous

• regenerate knowledge index after merging #1166-#1169 (cbd3dda)
• reports: publish delegation-economics orchestration benchmark (#1099) (8d37283)

10.12.2 (2026-07-18)

Bug Fixes

• route plan-review personas by the medium effort band, not a hardcoded sonnet (97f4943)
• trim orchestrator.md effort-band guidance to rubric-only and guard against agent-name drift (edd1519)

10.12.1 (2026-07-18)

Bug Fixes

• load plugin hooks via hooks/hooks.json and route subagent models by dispatch alias (2bd06f3), closes #1178

10.12.0 (2026-07-18)

Features

• hooks: per-agent-type cost attribution in the cost meter (cb5fb7b)
• skills: add orchestration-benchmark skill — pre-registered solo vs coordinated A/B protocol (8e8776f)

Documentation

• adr: record Claude-only model-routing boundary as a deliberate scope decision (8cb3f85), closes #1096

Miscellaneous

• regenerate knowledge and skills indexes for Phase 0 instrumentation (d70b6a8)

10.11.0 (2026-07-17)

Features

• build: default parallel build concurrency to wave width bounded by cores (a331cf0), closes #1170

10.10.0 (2026-07-17)

Features

• mutation-testing: detect and classify xunit.v3 shim-breaking test constructs (8908cd4)

Bug Fixes

• mutation-kill: add shim-first feasibility gate before the loop (a2e4e01)
• mutation-testing: harden the v2 shim path and add the -t mtp coverage-off floor (b3427a3)

10.9.1 (2026-07-17)

Bug Fixes

• mutation-testing: detect and surface Stryker coverage-capture failure (8821f77)

10.9.0 (2026-07-17)

Features

• test-improve: --from-phase auto-detects resume point when no number given (a8fe39f), closes #1151

Bug Fixes

• project-init: split Step 4c so keyless code-lookup pair decouples from key-gated Graphify (88facec), closes #1141
• setup: add --legacy-peer-deps fallback for npm install ERESOLVE (9e6ae89), closes #1142
• setup: stack-gate Step 6 mutation-tool install so wrong-stack probes can't fire (93ecc7a), closes #1152
• setup: write stryker.config.mjs directly instead of interactive init (f16e662), closes #1140
• test-improve: gate refactor steps on no-refactor mode at Phase 4b/5 (9146cb5), closes #1146

10.8.0 (2026-07-16)

Features

• mutation-testing: add --headless claude --print generation to the kill loop (32d4c47)
• mutation-testing: add config-driven survivor-kill loop mechanics (82e6203)
• mutation-testing: add generic Stryker shard-config generator (d81d98b)
• mutation-testing: add honest-score mutation report module (8b0b500)
• mutation-testing: add sharded worktree pipeline with timeout-abort (03d0e7e)
• reports: Slice 1 — Shared render module + bundled print stylesheet (d1addba)
• reports: Slice 2 — Standalone /report-pdf command (ccd1a3f)
• reports: Slice 3 — --pdf flag on the five report-producing skills (afdeab5), closes #1114

Bug Fixes

• mutation-testing: harden shard-setup path, bound streaming abort, dedupe report parsing (5eda4bf)
• mutation-testing: stop pinning a model snapshot id in the kill loop (fe2a76e)

Code Refactoring

• mutation-testing: point mutation-kill agent at the migrated scripts (2b90c44)

10.7.0 (2026-07-16)

Features

• /ship resume guard detects already-shipped or in-flight issues (b04c8f7), closes #1121
• all-or-none code-lookup tool install + non-review agent→tool mapping (17f9451), closes #1108
• code-review: consolidate sliced findings with dedup and theme rollup (fd37367)
• code-review: partition engine and activation policy for sliced review (85f1d41)
• code-review: persist per-slice findings with progress ledger (b2c8f58)
• code-review: reduced review panel for declarative slices (1b1d816)
• code-review: resume interrupted sliced runs (10a9cb5)
• install + build code-intelligence indexes in setup, disclose graphify key cost (617c5a3), closes #1134 #1135
• test-improve: three-way mutation mode and opt-in baseline-metrics report (dcf5d30), closes #1126

Bug Fixes

• /setup gitignores dev-team runtime artifact folders in target repos (8f98fc7), closes #1101
• allow full pytest suite to run in one invocation (eaaea0a), closes #1120
• grant code-intelligence MCP tools to read-only review agents (87d1895), closes #1102
• resolve agent knowledge-file paths via CLAUDE_PLUGIN_ROOT (#1103) (3118f80)
• scope /pr code-review gate to branch diff instead of full repo (3176cf5), closes #1122

Documentation

• code-review: tighten sliced-mode docs and section-artifact schema (11e9776)

10.6.2 (2026-07-15)

Bug Fixes

• check Vitest coverage provider in coverage-baseline readiness (a229d0a), closes #1089

10.6.1 (2026-07-15)

Bug Fixes

• verify coverage-baseline readiness during /setup for JS/TS repos (f2a8240), closes #1086

10.6.0 (2026-07-14)

Features

• add stryker-xunit-v2-shim skill and hook for xunit.v3 mutation scoring (fcd9e0f), closes #1083

10.5.0 (2026-07-10)

Features

• help: show curated main workflows by default, --all for everything (c825c23)

Bug Fixes

• help: regenerate knowledge/index.json after SKILL.md edit (b52129c)
• resolve Python 3 across interpreter names so hooks and /version work on Windows (e8fe721), closes #1078

10.4.0 (2026-07-08)

Features

• add ai-provenance-review agent for AI-authored test verification debt (99b2545)
• add DEV_TEAM_REPORTS/ to project-init's gitignore template (#961) (b45de47)
• add oracle provenance audit and unarmored-region detection to test-review (96a5fe7)
• add prompt-injection defense and falsifiability anchors to adversarial-review-protocol (6ae2e90)
• add React, Vue, Angular reactivity-review agents (closes #1019) (eac1ff3)
• add zero-findings anomaly handling and tolerated-deviation hunt (db6af77)
• autoship: add discovery CLI with required cap validation and shared input seam (3c8ddda)
• autoship: add dry-run preview, live comment/relabel, and per-issue status reporting (9eeec6b)
• autoship: add pure eligibility filter, oldest-first cap, and stdout contract (f2c65d0)
• autoship: add pure orphan-staleness selector with --stale-after-hours flag (92b773d)
• autoship: add shared --input-file/--now-override CLI seam (cd0d73c)
• autoship: add shared round-timestamp and staleness helper functions (487e01b)
• autoship: wire live gh fetch (issue list + timeline label event) for reclaim (7915e9d)
• autoship: wire live gh issue list fetch with clear failure surfacing (53a3c2e)
• correctness-review recall gaps on trailing-decimal validation and missing guard clauses (bd11509)
• extend Context needs vocabulary with artifact-stream value (af057bc)
• generalize build_skills_index.py with --plugin-dir and commands/ support (3843b91)
• hooks: add task_completion_metrics.py Stop hook for JSONL audit logging (ead0dbf), closes #1044
• implement open issues batch (autoship, co-evolution, nav, audit) (c57c533), closes #1065 #1066 #1067 #1068 #1069 #1070 #1071
• migrate /triage to DEV_TEAM_REPORTS/triage/ (#961) (f829e1f)
• pass --internal on /build's review-agent dispatches (#961) (197f20a)
• plan: accept a supplied --spec-issue as the spec source (562ce90)
• point 3 report-writing skills at shared report-template.md (#969) (6c251ed)
• preserve Claude Code OAuth login for isolated headless dispatch (e0bdf9d)
• specs: add tested opt-in-marker detection helper for issue-first specs convention (9c58861)
• specs: persist to a GitHub issue on opted-in, GitHub-connected repos (bfff41d)
• test-improve: add post-Phase-7 re-run-with-refactor close-out prompt (de52495)
• test-improve: add Tests-by-type table to executive-summary template (#962) (792132f)
• test-improve: foreground REFACTOR_REQUIRED items in Phase 7's §7 table (0d88af2)
• test-improve: persist test-counts-after.json via identical Phase 6 classification pass (#962) (3e6c668)
• test-improve: persist test-counts-before.json via Phase 1 classification pass (#962) (29f37d2)
• test-improve: suggest /handoff after context-heavy phase boundaries (548208b)
• test-improve: thread --refactor-mode into /quality-targets-converge's dispatch table (dbd2582)
• test-review: detect reflection-based private-member access per language (#959) (070bd42)
• write DEV_TEAM_REPORTS/<agent>.md for human-invoked /review-agent (#961) (a9b4a14)
• write DEV_TEAM_REPORTS/code-review.md unless --json or --internal (#961) (6017b33)

Bug Fixes

• add /test-improve to /code-review Step 6's non-interactive exception list (7a0270e), closes #995
• add ai-provenance-review calibration floor entry for CI (e6a531b)
• add ai-provenance-review to docs/agent_info.md Review Agents table (bf7d5a8)
• add anchor fragments to secondary reactive-effect-patterns references (fe6512d)
• add anchor fragments to test-review knowledge citations (147cd07)
• add Whole-file load tokens and agent_info rows for reactivity review agents (2ca9a55)
• autoship: address /code-review backstop findings for #989 (42569b9)
• autoship: address slice-boundary review findings for #989 (248079a)
• autoship: catch missing gh binary and distinguish error sources for #989 (3503c4b)
• bump ui-ux-designer effort band to high (e3f866b), closes #1013
• copy_auth_state() must copy ~/.claude/, not just ~/.claude.json (6226da2)
• correct broken skills/test-improve path reference in report-template.md (#969) (b012c8c)
• grant Write in review-agent and /review allowed-tools for DEV_TEAM_REPORTS writes (#961) (6694d49)
• guard /pr against accidental issue auto-close via negated closing keywords (228aae8), closes #977
• harden /code-review --json instruction against narration-instead-of-emission (cebff56)
• make competitive-analysis reference sentence byte-identical (#969) (90ec407)
• mention /test-improve alongside /ship in workflows doc (bab5fcb)
• mention /test-improve alongside /ship in workflows doc (2ac5293)
• pass --internal on /test-improve's internal /code-review dispatches (#982) (618e3af)
• qualify competitive-analysis classification labels with (dev-team) (3419b63), closes #1016
• regenerate knowledge index (27af4df)
• regenerate knowledge index and add eval fixtures for new reactivity review agents (7888a2b)
• regenerate knowledge index for adversarial-review-protocol new section (3ac3159)
• regenerate knowledge index for CI (05200dd)
• regenerate knowledge index for oracle-provenance.md (92f3441)
• restore required claims-discipline phrase in CLAUDE.md (dd618e7)
• retry once via --resume when /code-review --json narrates instead of emitting JSON (e69f806)
• test-improve: address code-review findings for #968 (c98aec8)
• test-improve: correct --path reference and tie-break contradiction; update consumer docs (#962) (8bf411e)
• trim CLAUDE.md to under 5000 chars for CI gate (31ea276)

Code Refactoring

• self-declared review-agent dispatch scope via frontmatter (aaef09d)
• test-improve: trim SKILL.md under the 500-line token-efficiency guard (62d238c)

Documentation

• add --internal to review-agent's argument-hint (#961) (8839dce)
• add report-output-location.md (DEV_TEAM_REPORTS/ convention) (#961) (c1e2a4b)
• add report-to-pdf.md rendering recipe (#969) (355ba09)
• add shared report-template.md contract (#969) (98e85ed)
• address code-review findings for #986 (6fdbb90)
• clarify report-template.md adoption status and align competitive-analysis header fields (#969) (35e3eba)
• cross-reference report-output-location.md from the three migrated skills (#961) (16e0c6e)
• documentation overhaul for all three plugins (d2e5caf)
• fix C#/Reqnroll wiring instructions in bdd-frameworks.md (3c0800b), closes #948
• move feedback-learning keyword detection out of orchestrator routing (bdf517e)
• name /test-improve's --internal caller status and /ship's exception (#982) (abaefb0)
• recommend -t mtp default for Stryker.NET on .NET 10 targets (da9fbda), closes #947
• sharpen /ship exception rationale and fix step-6/7 mislabel (#982) (3b30712)
• test-improve: clarify static_analysis counting and update Phase 7 file list (#962) (2c13a9b)
• trim redundant --internal/--json rationale in build/SKILL.md (#961) (2a5ac25)
• update triage cross-references to DEV_TEAM_REPORTS/triage/ (#961) (58bfe9b)

Miscellaneous

• rebuild knowledge index after build/SKILL.md trim (#961) (8188c35)
• rebuild knowledge index after DEV_TEAM_REPORTS/ migration (#961) (091d622)
• rebuild knowledge index after issue #982 audit (#982) (882a782)
• rebuild knowledge index after report-template edits (#969) (323883e)
• regenerate docs/skills.md after triage migration (#961) (22d9023)
• remove dead .sh byte-parity test cases (#976) (dea8c68)
• specs: rebuild knowledge index for #986; trim plan/SKILL.md to fit line-count guard (0c10cc7)
• test-improve: add Enforced-by header, rebuild knowledge index for #968 (5ab2af3)

10.3.2 (2026-07-06)

Bug Fixes

• add unconfirmed outcome to /triage's root-cause schema (460f649), closes #918 #924 #925 #926 #927
• remove test-health pain-point calibration question (c5d8f7c), closes #939

10.3.1 (2026-07-06)

Bug Fixes

• add explicit fallback for degenerate branch-base resolution in /build Step 7 (744d321), closes #916
• auto-detect src-layout Python projects for mypy invocation (c35cbbc), closes #917
• clarify slice-checkbox flip must wait for verify and invariants (7f9c33b), closes #915
• detect Python test files in is_test_file() (3bd7090), closes #913
• evaluate all bash write patterns instead of first-match-wins (e107114), closes #914

Miscellaneous

• merge /init-dev-team into /setup as a single provisioning command (a5086c3), closes #937

10.3.0 (2026-07-06)

⚠ BREAKING CHANGES

• the user-invocable /context-summarization command no longer exists; use /handoff instead. Chose an outright rename over a deprecated alias because this is a young, actively-developed plugin with no external consumers depending on command stability across releases, and the two skills share their entire mechanic (forget/ input/output gates writing structured markdown for another session) -- a thin alias would mean carrying dead command-name plumbing for a release cycle for no real backward-compat benefit.

Features

• add --ablation mode to /agent-eval for causal drop-candidate evidence (3463958), closes #868
• add --calibrate band ladder mode to /agent-eval (b7f35cd)
• add /harness-e2e-check as a repeatable on-demand integration test (4c0099a)
• add boundary-events telemetry channel and intervention events (343a4a8), closes #859
• add branch/remote-aware risk escalation to destructive_guard (a0e4912), closes #862
• add calibration floors policy table and guard test (39d6bca)
• add correctness-review agent for functional-defect detection (093cb7b), closes #885
• add deterministic failure-class routing to build and apply-fixes repair loops (30969ed), closes #861
• add failure-signature dead-end detection to /build repair loop (19cfa87), closes #864
• add invariants, rollback points, and declared file scope to plan slices (b33d561), closes #865
• add per-edit authoring-discipline checklist to software-engineer agent (6f8fe5c), closes #886
• add preventive Bash-aware test-freeze guard and boundary-events wiring for revert guard (9bf96c3)
• add proposed/evidence/risk fields to gate-decision audit records (0da7a70), closes #867
• add recalibration staleness advisory to /model-routing-check (11964d3)
• add shared directory-enumeration guidance for existence-checking agents (030237d), closes #878
• add structured evidence bundles to build output and PR bodies (16887d1), closes #863
• add validated-outcome weighting to feedback-learning lessons (e37d871)
• gate feedback-learning agent mutations on /agent-eval change contracts (7d4554b)
• rename context-summarization skill to handoff, add fork mode (941e25d), closes #853

Bug Fixes

• add missing Role declarations to /specs, /design-doc, and workflow skills (03ab5bd), closes #884
• correct misrouted watchlist.md reference in skills-registry.md (056054b)
• reclassify correctness-review to high-risk calibration floor (b491625)

10.2.1 (2026-07-05)

Bug Fixes

• guard claude-setup-review against Read-on-directory EISDIR (49f0075)
• harden code-review against EISDIR on --path dirs and json non-determinism (a3dce5f)

10.2.0 (2026-07-05)

Features

• project-init: wip graph-tools consolidation (codegraph + graphify) (543f2b2)

Bug Fixes

• harden cloud plugin freshness (scope-aware update, drift advisory, snippet auto-update) (43fea0a)

Code Refactoring

• complete Classic TDD cadence removal; codify defect-fix test-first rule (6dba0d4)
• de-duplicate /setup vs /project-init; expand /init-dev-team scope (93a4345)
• remove Classic TDD opt-in cadence; wire graph guidance into recon/architect (93669b0)
• single source of truth for the plugin auto-update flag (cfaaf18)

Documentation

• add codegraph-vs-graphify comparison; wip build cadence cleanup (02900fb)
• align command taxonomy and developer-notes index with the graph-tools step (3d873a3)

Miscellaneous

• regenerate knowledge index for the /upgrade skill edit (b51dc98)
• regenerate knowledge/index.json (24e2601)
• regenerate skills index; fix knowledge-anchor citations (9b287eb)

10.1.0 (2026-07-05)

Features

• add /headless-run skill for isolated headless claude -p dispatch (33e857d), closes #842
• project-init: install capability tools beyond the static-analysis lanes (e68ebd9), closes #838

Bug Fixes

• give adr-author Bash and Skill so it drives the adr-tools CLI (715fa39), closes #837
• harden the /code-review skill's tools, output paths, and --json contract (c979587), closes #834 #835 #836 #841 #843
• point cross-skill capability-tools references at the plugin-root path (180e644), closes #838
• route missing-tool commands to the onboarding command (4c31199), closes #838
• wire design-it-twice into the architect agent's skill dispatch (43af8c8), closes #833

Miscellaneous

• regenerate knowledge and skills indexes (2ec8ed7)

10.0.0 (2026-07-05)

⚠ BREAKING CHANGES

• the /js-project-init slash command is removed; use /project-init. The old natural-language trigger phrases still resolve via the renamed skill's description.

Features

• add Python static-analysis lane (Ruff + mypy) (e8d9de3), closes #807
• build: code-first implementer cycle; forbid test edits during refactor (#813) (6a41b10)
• build: default cadence is Code-First Small Batches; TDD opt-in (#813) (4d3331f)
• build: run static self-heal mechanism at review checkpoints (#811) (a3dd9ba)
• code-review: dispatch component-architecture-review on UI components (fb5c47b)
• generalize js-project-init into project-init (#822) (887062e)
• hooks: block test-file edits during refactor phase with recovery guidance (#813) (7be2569)
• hooks: register tests-frozen freeze and revert guards (#813) (974afff)
• hooks: revert Bash-mediated test edits during refactor phase (#813) (f0b1d89)
• hooks: shared test-file classifier + phase-state reader (#813) (207d9a0)
• orchestrator: fast-path single-module standard tasks per Rec 2 (#813) (6d9cc54)
• plan: cadence metadata + code-first step template (#813) (aba40ab)
• static-analysis: add build-time lane registry and language-setup skeletons (#811) (2bae49c)
• static-analysis: add Java lane with PMD diagnostic and Tier 1 SARIF source (253944e), closes #810
• static-analysis: register C# lane via dotnet format and Roslyn ErrorLog SARIF (0266fe0), closes #809
• static-analysis: register oxlint as the JS/TS lane and /code-review source (#808) (ebed57e)

Bug Fixes

• plan: cite 01-final-results.md instead of retired consolidated report (#813) (6e17cdd)
• surface headless assumptions and escalations to the PR (#825) (e893f43)

Documentation

• add Developer Notes page with add-a-language playbook (#812) (e38c438)
• cadence-claim sweep for code-first default; rebuild knowledge index (#813) (7af590d)
• code-review: review agents are the quality gate; coverage/mutation saturate (#813) (c7dc66d)
• document the /triage → corrections → /apply-fixes workflow end to end (b7dbd43), closes #819
• drop issue refs and inline rationale from CLAUDE.md rules (21199b7)
• move experiment pre-registrations and data into an evidence folder (e2633fa)
• specs: state ambiguity-resolution positioning per Rec 1 (#813) (8d5d15d)

Miscellaneous

• gitignore plans/ — plan drafts stay untracked like docs/specs (9e2b416)
• regenerate skills index after specs description update (#813) (524a905)
• retire tracked working documents; require npm ci in new worktrees (59e3867)

9.3.0 (2026-07-04)

Features

• add BDD feature-file scan to detect_bdd_convention (#537) (6067c1d)
• add plan_gherkin_export for byte-for-byte .feature persistence (#537) (5a70a14)
• decision-aware no-op modes and overwrite reporting in plan_gherkin_export (#537) (1f30212)
• detect BDD convention and record persistence decision in /plan (#537) (d3f87fc)
• export approved plans' Gherkin via plan_gherkin_export post-approval (#537) (1344fa6)
• failure-path CLI contract for plan_gherkin_export (#537) (4f79afe)
• graduate context ceiling warning to summarization action bands (29f5c3a)
• hooks: add absolute-token cap to context ceiling guard (0caa619), closes #786
• hooks: auto-detect context window from transcript model (d1db9bd), closes #785
• hooks: graduated bands keyed to multiples of the effective ceiling (0abf2b0), closes #781
• hooks: pin context-window detection to specific 1M model versions (be80f77), closes #779
• hooks: pinned message contract with bound + window provenance (2be8fe2), closes #780
• JSON CLI contract for detect_bdd_convention (#537) (e770279)
• manifest signals and precedence in detect_bdd_convention (#537) (85a6c61)
• record Gherkin persistence decision in plan template (#537) (61a5ff6)

Bug Fixes

• byte-level export fidelity and manifest-matcher hardening from slice reviews (#537) (4c36c6a)
• drop pinned model snapshot id from plan SKILL.md example (#778) (32e11f2)
• harden plan_gherkin_export against escaping destinations (#537) (21b664d)
• invoke shipped python scripts with python3, not bash (#777) (82c63ea)
• never write .feature exports through planted symlinks (#537) (ef138cf)
• pin directory-only persistence recording and triad guard (#537) (af0fda2)
• point install link and SARIF informationUri at bdfinst/agentic-dev-team (#805) (567652e)
• restore exact claims-discipline wording in CLAUDE.md (830677c)
• tier classification no longer over-triggers on the /ship-required default stance (#778) (6cbf2b8)

Code Refactoring

• extract Gherkin persistence procedure to a plan reference file (#537) (c5c8a65)

Documentation

• add Context Management user guide (f9c7c70), closes #783
• align context ceiling docs with auto-detected window, 150K cap, and evidence-based 40% rationale (f714eae)
• align skills, CLAUDE.md, and ADR 0011 with the shipped guard (42fa885), closes #782
• document model id-to-tier mapping for plan-review persona dispatch (#778) (83ec7a4)
• note Gherkin .feature persistence in the /plan registry entry (#537) (5ebf2e3)
• trim CLAUDE.md back under the 5000-char budget (c4559be)

Miscellaneous

• declare bdd-frameworks.md in slice 1 file surface (#537) (d2c12d5)
• rebuild knowledge index for context-ceiling doc updates (c44ea8f)

9.2.0 (2026-07-02)

Features

• extend check_md_references.py to validate anchor fragments (f7dd077)

Bug Fixes

• build: mandate /verify before a runtime-surface slice is done (1ca3665)
• claude-md: restore claims-discipline sensor phrase after trim (91ae4c5)
• docs: use fully-qualified dev-team:tech-writer in dispatch prose (f0ca139), closes #720
• hooks: escalate verify-guard to a hard block for stuck loops (3fd772c), closes #708
• hooks: purge stale bash-retry/verify-guard state files on write (66cf777)
• hooks: replace third-party PyYAML with a minimal stdlib parser (bba1f9b)
• hooks: require a logged reason for --no-verify/-n commit bypass (4e3470c), closes #709
• hooks: use in-process hashing in verify_guard_state instead of cksum (08ee6e0)
• install: retire Git Bash hard-requirement on Windows (5704a04)
• model-routing-check: invoke resolver with python3, not bash (340f14c), closes #718
• mutation-testing: avoid real OS signal delivery in portable test (f11beaa)
• mutation-testing: terminate all live Stryker procs on signal, not one (6de8d95)
• proxy-resilience: consolidate proxy pointers to stay under CLAUDE.md budget (4943e4e)
• scripts: bound shell subprocess calls with a timeout (cb5e802)
• skills: add diagnose-before-retry rule to /build (c896654), closes #710
• skills: add pre-switch verification to branch-workflow (4be5516), closes #714
• skills: extend quality-gate-pipeline with external UI citation guidance (91746bb), closes #728
• skills: pair branch-workflow cleanup with EnterWorktree/ExitWorktree (3751720), closes #716
• telemetry: resolve dead-surface audit of never-invoked agents (#712) (9b8a7c2)
• tests: extend hermetic-adoption gate to plugins/dev-team/tests/*/.py (649c4ce), closes #715

Code Refactoring

• hooks: dedupe model-band helpers into model_resolve (758a28b)
• hooks: dedupe stdin-JSON-read helper into hooks/lib/stdin_json.py (dbfb953)
• hooks: rename magic numbers and cryptic identifiers in 4 hooks (b3d8d07)
• hooks: share the 5000-char / 500-line token-efficiency thresholds (7a44221)
• hooks: stop double-parsing the ladder JSON in resolve_band() (b73a444)
• mutation-adapters: dedupe first-changed-source-file lookup (935ad37)
• scripts: dedupe issue-dict builders into review_result.make_issue (ed212e1)
• skills: extract plan file template to references/plan-template.md (dad1b07)

Performance Improvements

• hooks: rotate disposed entries out of pending-review.jsonl (12c9a41)
• hooks: tail-read cost meter transcript instead of full re-parse (0e73d8c)

Documentation

• agents,skills: recover from Edit stale-old_string via re-Read (f5fa055), closes #726
• agents: sync agent_info.md review-agent table with agent-registry.md (1e5dadd)
• ci-debugging: add Monitor timeout_ms guidance for long-running jobs (a814a67), closes #721
• document frontmatter extension keys, drop stale skip-guard, clarify ADR 0015 scope (680c060)
• fix bash-migration doc drift (dead agent refs, stale .sh names, ADR TOC) (c991c3d)
• knowledge: document proxy connection-refused retry/surface policy (7607bba), closes #724
• proxy-resilience: add backoff/retry ceiling convention for proxy rate limits (f3ded18), closes #723
• skills: add fast feedback loop guidance to browser-testing (7fe8b43), closes #725

9.1.0 (2026-07-02)

Features

• mutation-testing: first-class Stryker.NET slicing + slice-level parallelism (#561) (1cc52b0)

Bug Fixes

• skills: grant code-review the Agent tool it already instructs (37568f4), closes #730

9.0.0 (2026-07-02)

⚠ BREAKING CHANGES

• mutation-testing: The shipped bash scripts (csharp-stryker-net-wrapper.sh, csharp-stryker-net-status-loop.sh) are removed. Downstream .NET operators who copied them into their repo's scripts/ must migrate to csharp_stryker_net_wrapper.py + csharp_stryker_net_status_loop.py. See plugins/dev-team/skills/mutation-testing/references/languages/csharp-stryker-net.md for the current install + run instructions.
• test-improve: consolidate /test-modernize and /test-upgrade into /test-improve (#536) (#566)

Features

• agents: mutation-kill catches DI-wiring files by signal alone (#667) (1206210)
• agents: mutation-kill escalation respects the Standard CompileError trap (#667) (ff470c5)
• agents: mutation-kill persists per-file convergence history (#667) (b072919)
• agents: mutation-kill shrinks --mutate glob from convergence history (#667) (b71fd9a)
• agents: mutation-kill tiers Basic-then-Standard mutation levels (#667) (8c25d33)
• hooks: port agent-model-resolve to python (#585) (e9c9396), closes #572
• hooks: port bash-retry-guard to python (#591) (#629) (4ea68d8), closes #572
• hooks: port codegraph-bootstrap to python (#592) (#632) (189ac4d), closes #572
• hooks: port codegraph-nudge to python (#593) (#634) (9507ddf), closes #572
• hooks: port codegraph-turn-mark to python (#594) (#638) (f3bbb4d), closes #572
• hooks: port context-ceiling-guard to python (#595) (#644) (05a57dd), closes #572
• hooks: port contract-version-guard to python (#596) (dffa130), closes #572
• hooks: port cost-meter dispatcher to python (#597) (f6859b5), closes #572
• hooks: port destructive-guard to python (#598) (#654) (1f94854), closes #572
• hooks: port eval-compliance-check to python (#599) (e1ed65d), closes #572
• hooks: port js-fp-review to python (#600) (bab8507), closes #572
• hooks: port knowledge-index hook to python (#580) (#639) (fb15a9d), closes #572
• hooks: port knowledge-index-paths shared lib to python (#575) (#621) (ff1ebbb), closes #572
• hooks: port pre-commit-detect + review-gate-hash shared libs to python (#576) (#623) (7dd297a), closes #572
• hooks: port pre-commit-knowledge-index to python (#582) (#643) (2c98625), closes #572
• hooks: port pre-commit-review to python (#583) (#646) (42c47bd), closes #572
• hooks: port session-model-banner to python (#586) (5c4cb5e), closes #572
• hooks: python hook contract + parity harness + reference port (#574) (#620) (b1ccc82), closes #572
• hooks: python port of token-efficiency-review (#607) (#647) (db40b79), closes #572
• hooks: python port of verify-guard (#608) (#651) (2b2a461), closes #572
• hooks: python port of version-check (#609) (#628) (ec80ed3), closes #572
• install: port install.sh to python + shell trampoline (#616) (d5ed5fc), closes #572
• mutation-testing: pretooluse hook enforces step 1c smoke gate (#565) (#569) (2b62b50)
• mutation-testing: python port of wrapper + status loop (#572 phase 1) (#573) (620e8d8)
• mutation-testing: reference wrapper + status loop for .NET long runs (closes #558, #559) (#563) (e3e4b2f)
• mutation-testing: workflow-enforced smoke gate + solution-path warning (closes #554, #557) (#562) (3151e2d)
• scripts: --stryker-concurrency/STRYKER_MUTANT_CONCURRENCY override for wrapper default (#667) (8a53dd1)
• scripts: port build-worktree-baseref to python (#587) (eda5527), closes #572
• scripts: python port of build-jobs (#610) (#642) (77ceb14), closes #572
• scripts: python port of build-wave (#611) (#631) (13fbc3e), closes #572
• scripts: python port of build-wave-reconcile (#612) (#637) (bd57b1d), closes #572
• scripts: python port of git-origin-host (#613) (#622) (3b2f0d8), closes #572
• scripts: python port of issue-deps (#614) (#653) (e3e26f0), closes #572
• scripts: python port of recon-inventory (#615) (6c956ac), closes #572
• scripts: stryker.net wrapper defaults concurrency to cores-2 (#667) (ab4ca3f)
• skills: python port of collect-domain-signals (#617) (#657) (585a755), closes #572
• test-improve: consolidate /test-modernize and /test-upgrade into /test-improve (#536) (#566) (d6563f5)
• test-improve: scaffold + worker parameterization (Slices 1, 11 of #536) (#555) (d8a6ad1)

Bug Fixes

• agents: correct quality-targets-converge/SKILL.md cross-reference path (1f2faf2)
• build: worktree agents inherit caller HEAD via detect-and-warn (#553) (#560) (2655c83)
• ci: cross-platform pytest fixtures for the wrapper on windows (#619) (4dfe28f)
• mutation-testing: probe dotnet-root across macos + linux + windows git bash (#564) (#568) (143bbd9)

Code Refactoring

• mutation-testing: remove bash wrapper + tests, adopt python-only (#581) (f2fb9e9)
• test-design: resolve smell-review/advisor remedy overlap at source (#534) (#547) (e31b09c)

Documentation

• mutation-testing: document convergence-derived glob negations (#667) (045dc5f)
• mutation-testing: document tiered mutation-level example commands (#667) (03b79db)
• mutation-testing: document wrapper concurrency default (#667) (9061fc6)
• mutation-testing: fix cross-reference direction and soften overclaimed parity wording (6927c59)
• mutation-testing: fix intra-file anchor slug for the new perTest section (d2000d1)
• mutation-testing: recommend coverage-analysis perTest default for xunit.v2-shim Stryker.NET projects (0770c27)
• mutation-testing: recommend local install with language-specific install commands (#556) (d277694)
• mutation-testing: stop showing coverage-analysis as a Stryker.NET CLI flag (47e8825)
• mutation-testing: warn that '<tool> | tee' masks exit code (#550) (#552) (b65e90e)
• update cross-references to bats files ported in #671 (c8ed195)

Miscellaneous

• 572: remove all shipped bash + parity harness; close epic (#618) (d6af7e2)
• hooks: port model-resolve shared lib to python (#572.C0) (#626) (9d6715d), closes #577
• hooks: port mutation-adapters lib + 4 adapters to python (#572.D0) (#625) (4308291), closes #578
• hooks: port mutation-gate to python (#572) (#655) (b0681b7), closes #588
• hooks: port post-format to python (#572) (#630) (b6bae0c), closes #601
• hooks: port pre-tool-guard to python (#572) (#633) (ee51a0c), closes #602
• hooks: port session-learning-trigger to python (#572) (#636) (b0715c3), closes #603
• hooks: port skills-index dispatcher to python (#572) (#641) (c65b096), closes #604
• hooks: port tdd-guard to python (#572) (#645) (8cf5bb4), closes #605
• hooks: port telemetry dispatcher to python (#572) (#648) (52f411a), closes #606
• hooks: remove mutation-testing-smoke-gate.sh; python default (#584) (95404df)
• remove ACI-branded references from shipped plugin content (98a604c), closes #678
• scripts: port plan-parse shared lib to python (#572.E0) (#627) (c9c4e57), closes #579
• scripts: port plan-waves to python (#572) (#649) (3e88ff5), closes #589

8.4.0 (2026-07-01)

Features

• mutation-testing: 8 improvements from .NET mutation drive (#544) (6648b80)
• mutation-testing: honest score, timeout warning, NoCoverage-first triage, (#521) (#545) (069cd00)
• mutmut: add Python mutation testing adapter + dispatch wiring (#518) (69da27f)
• stack-aware: wire reference loading into three test-strategy skills/agents (#524) (#530) (2f97705)

Bug Fixes

• knowledge: correct misrouted csharp-http-client-testing.md references (#520) (0a3a70d)
• pitest: class scoping, withHistory, per-mutant timeout, multi-module Maven (#517) (e85ae92)
• plan: remove AC mirror from Build Progress template (#526) (#538) (f60df58)
• stryker-js: raise timeout 60s→300s, read per-mutant timeoutMS from config (#516) (6dce929)
• stryker-net: shard-aware execution prevents mutation gate timeouts on large C# repos (#515) (c8f6624)
• test-design: scope farley score to --path and --since (#533) (#542) (00925d8)

Code Refactoring

• mutation-testing: split SKILL into language-agnostic workflow + per-language KBs (#523) (9ae29e2)
• test-design: demote test-design-advisor from user-invocable (#532) (#539) (5b3782b)

Documentation

• mutation-testing: correct Stryker.NET reference (xunit.v3, DOTNET_ROOT, CLI, verbosity) (#540) (d9c0f1c), closes #522

8.3.4 (2026-06-29)

Documentation

• group the skills catalog by capability, not invocation type (#511) (ea87e20)

8.3.3 (2026-06-29)

Documentation

• render mermaid in MkDocs and unify the diagram strategy (#508) (dc32a63)

8.3.2 (2026-06-29)

Documentation

• improve findability and trim verbosity (#502) (fade9e0)
• simplify language to high school reading level (#506) (3dc30bf)

8.3.1 (2026-06-28)

Documentation

• correct three-phase workflow diagram to match actual workflow (#498) (bde853f)
• fix diagram rendering in dark mode and resolve overlaps/clipping (#496) (03bbe36)

8.3.0 (2026-06-28)

Features

• add frontend component-architecture review agent and /frontend-architecture skill (#492) (4f459bc)

8.2.0 (2026-06-27)

Features

• guard markdown references and harden skill runtime reads (#489) (184725f)

8.1.0 (2026-06-27)

Features

• eval fixtures for uncovered review agents; plugin-audit skill fixes (#485) (6ede1cf)

8.0.0 (2026-06-26)

⚠ BREAKING CHANGES

• the agent-create, agent-skill-authoring, agent-add, agent-remove, and add-plugin skills are removed from dev-team (hard cut) and now live in marketplace-dev. dev-team registry, CLAUDE.md, docs, orchestrator, tech-writer, and knowledge index updated accordingly. The agent-create effort test is repointed to the migrated skills. agent-audit remains in dev-team.

Features

• add /gherkin-derive skill (standalone Gherkin derivation from code) (#483) (fca5d63), closes #441
• add /test-upgrade skill (general-purpose analyze-then-improve orchestrator) (#484) (2c03350), closes #442 #443
• add Cucumber-JVM BDD entry to Spring Boot test stack profile (#480) (1cfb759), closes #440
• add Cucumber.js BDD entry to Node.js test stack profile (#479) (1228136), closes #439
• add Go mutation testing support to tool-setup and mutation-testing (#474) (5e8923d), closes #434
• add Godog BDD entry to Go test stack profile (#477) (98a24c1), closes #437
• add LOW_VALUE gap classification to test-health, specs, plan, and test-smell-review (#471) (d25329f), closes #431
• add marketplace-dev plugin and migrate plugin-authoring skills from dev-team (#464) (770e386)
• add mutation-kill agent (autonomous survivor-reduction loop) (#482) (f58eaf7), closes #461
• add Reqnroll BDD entry to .NET test stack profile (#478) (fecead7), closes #438
• auto-bootstrap the CodeGraph index per clone (#428) (ba9ff78)
• convert six markdown agents to Python/hybrid harness pattern (#462) (f587109)
• create bdd-frameworks.md knowledge reference for per-language BDD wire-in (#476) (b22a675), closes #436
• create bdd-value-guide.md knowledge reference (#475) (da2e9ef), closes #435
• enforce the 40% context ceiling with a PreToolUse hook (#425) (7fac5b4)

Bug Fixes

• agent audit compliance fixes — session-analysis schema, Context needs, persona, hook reliability (#465) (7ce0a6a)
• decouple coverage-baseline from /test-modernize (#472) (085fca3), closes #432
• decouple coverage-delta from /test-modernize (#473) (f432452), closes #433

Documentation

• reduce plugin CLAUDE.md from 30k to 4950 chars via knowledge file extraction (#468) (561e457)

Miscellaneous

• trigger ship skill more broadly (#470) (fcb5924)

7.9.0 (2026-06-25)

Features

• deterministic status + finding grouping for doc/naming review agents (#423) (8b6bdbe)

7.8.0 (2026-06-25)

Features

• add craftsmanship-axis review rules (use-the-platform, comment hygiene) (#419) (50f761e)

Code Refactoring

• name shipped AC references for what they assert (#422) (f8c001d)

7.7.0 (2026-06-24)

Features

• add harness fixes from session-review and harness-audit (#398) (34c955a)
• add when-tdd-pays experiment fixtures and harness extension (#404) (4907f34)
• implement closed learning loop (closes #401) (#403) (d894ec4)

Bug Fixes

• add ambiguity resolution protocol to /specs and wire /ship gate (#412) (61c4b83)

7.6.0 (2026-06-23)

Features

• add Continuous Delivery knowledge from Humble & Farley (pipeline, release, data, maturity) (#390) (a91b7ed)
• extend DDD skills with supple design, distillation, and implicit-concept patterns from Evans (#388) (971f1d6)

7.5.0 (2026-06-23)

Features

• extend dev-team test tooling from xUnit Test Patterns + Working Effectively with Legacy Code (#386) (b5e2c27)

7.4.0 (2026-06-22)

Features

• run adversarial self-challenge in all 22 review agents (#382) (bd1a203)

7.3.0 (2026-06-22)

Features

• objective task-size classifier, no-plan fast path, and mutation timeout hardening (#379) (ae1e99c)
• quality-gate,harness-audit: trust review signals over saturating coverage metrics (#377) (9d22233)
• wire REFACTOR review loop, harden review agent determinism, add consistency evals (epic #362) (#378) (ec091a8)

7.2.0 (2026-06-22)

Features

• plan,build: tier plan review, batch inline review, headless approval gates (#355) (48c44e3)

7.1.0 (2026-06-22)

Features

• audit: add a registry-completeness sensor (scripts/check_registry_sync.py, surfaced as /agent-audit step 2e) that fails when an agent or skill on disk is missing from — or orphaned in — the registry tables (#350) (6a08144)

Code Refactoring

• agents: make the effort: frontmatter the single source of truth — remove the duplicate body Effort: line from every review agent and template, drop the hand-maintained Model Tier column from agent-registry.md, and have /agent-audit warn if the band is restated (#350) (6a08144)
• agents: raise the four judgment-driven reviewers (a11y, naming, complexity, progress-guardian) to effort: medium and reconcile the registry tier drift this surfaced (#350) (6a08144)

Documentation

• dev-team: note registry-completeness gate; release v7.1.0 (#352) (3630676)

7.0.0 (2026-06-21)

⚠ BREAKING CHANGES

• routing: agent frontmatter uses effort: bands, not model: tiers. The resolver accepts legacy model: tiers for one deprecation release.

Features

• dev-team: release ownership-engineering improvements; add commitlint guard (#339) (a689677)
• routing: effort-band model routing (replaces model: tiers) (#337) (2fda3c1)

6.11.1 (2026-06-20)

Code Refactoring

• test-review: consolidate overlaps, rename Farley scorer, dedupe shared prose (#324) (dd2d1da)

Documentation

• skills: disambiguate /agent-readiness from /harness-audit (#325) (f647247)

6.11.0 (2026-06-19)

Features

• eval: confidence-pyramid improvements (registry, dispatch, cache, citation lint, integration tier) (#315) (b560880)
• evals: backfill cites: frontmatter on reviewer agents (#319) (e37c627)
• evals: wire cache + integration tier + cites enforcement into /agent-eval and /agent-create (#322) (a3c623c)
• test-modernize: make Phase-3 disabled-test resolution a Phase-4 contract (#318) (e30ecba)
• test-modernize: per-Story mutation testing in Phase 4 + end-of-phase test review loop (MVP probe) (#316) (df30551)

6.10.1 (2026-06-19)

Documentation

• add /ship + /test-modernize workflow doc; alphabetize skill/agent tables (#277) (4eb563d)

Miscellaneous

• plugins: remove deprecated legacy stubs and rename-migration upgrade path (#275) (2048687)

6.10.0 (2026-06-18)

Features

• test-modernize: bind component tests to approved Gherkin scenarios (#273) (aa6bef7)

6.9.0 (2026-06-18)

Features

• skills: /test-modernize orchestrator workflow (#271) (3c81e0f)

6.8.0 (2026-06-18)

Features

• agents: add identity personas to all 11 team agents (#253) (cd751d9)
• build: JS project bootstrap gate — invoke js-project-init when package.json missing (#257) (4b6ff4e)
• js-project-init: add lint-staged with pre-commit auto-fix (#256) (d6d3c64)
• qa-engineer,test-design: rewrite qa-engineer as Senior SDET; lock test-design vocabulary to MinimumCD (#270) (f189ebf)
• version: make /version a mechanical, deterministic lookup (#259) (a553754)

Bug Fixes

• build: ship the parallel-build wave scripts inside the plugin (#261) (77d9e39)
• security-assessment: stop shipping build/test scripts; make runtime scripts discoverable (#263) (5ae7afc)
• upgrade: skip legacy-id migration when version >= 6.1.0 (#249) (770c413)

Performance Improvements

• ci: parallelize local pre-push gates and bats suites (#247) (836be51)

Documentation

• fix duplicate examples in GETTING-STARTED.md skill invocation section (#269) (0cc0c6a)

6.7.0 (2026-06-08)

Features

• build: wave-aware concurrent build (#224) (#242) (9137017)
• issues-from-plan: spec parent + DAG-linked slice children (#225) (#243) (3306499)
• plan: GitHub-origin post-plan issue gate (#226) (#244) (8d10fde)
• plan: parallelization-review persona (#223) (#241) (3a4a173)
• plan: slice dependency metadata + wave computation (#222) (#240) (e8c9aa9)
• security-scan: offline-harden gitleaks + trivy (#53) (#245) (0dd89ef)
• session-review: wire raw-log semantic tier + methodology lens (#214) (#239) (a1dd258)

Miscellaneous

• remove issue-tracked specs, spikes, and plans (#215) (#216) (d058860)

6.6.1 (2026-06-07)

Bug Fixes

• release: sync marketplace catalog via release-please extra-files (#210) (c84611a)

6.6.0 (2026-06-07)

Features

• agent-eval: eval variance aggregator — pass@k, flap, quarantine (#103) (#196) (4a2abcb)
• evals: incremental per-agent eval runs (not all-or-nothing) (#206) (8d7d9e0)
• evals: make the resumable sweep the default mode of run-full-eval.sh (#208) (bba7734)
• evals: resumable --sweep mode for run-full-eval.sh (#207) (088adce)
• evals: run-full-eval.sh — full corpus run + baseline refresh + auto-merge PR (#202) (37e0280)
• session-review: cross-machine union read (#178) + utilization fix (#182) (#188) (d6280b6)
• session-review: frequency→lever escalation (Delta C, #179) (#189) (3328c81)
• session-review: per-session gate instrumentation + bypass↔rework correlation (#111) (#200) (384138b)
• session-review: telemetry sync transport + config validation + security docs (#187) (559cf3d)
• telemetry: wire CI cost-regression gate to real cross-machine baseline (#171) (#192) (531a794)

Bug Fixes

• cost-meter: attribute by model + thread; drop inert buckets (#170) (#183) (c7a2b2a)
• mutation-gate: _timeout fallback must not write to stdout (#197) (12b1936)
• review-gate: bind .review-passed to staged CONTENT, not paths (#193) (#195) (5733af3)
• skills: reference prompt templates by explicit plugin-root path (#181) (a2be7cb), closes #173

Documentation

• adopt North Star + scope the self-improvement loop to /session-review deltas (#172) (6003609)
• concurrent-use: resolve #109 Phase 2 — one worktree per agent (#194) (3751b06)
• eval running guide, maintenance guide, and feature-verification plan (#201) (847fd02)
• how to give CI read-only access to the telemetry repo (#171 prep) (#191) (9fddacd)

Miscellaneous

• Wave 1 hygiene — scope honesty + orphan spec + beacon scope (#185) (86be934), closes #105 #115 #106

6.5.0 (2026-06-07)

Features

• cost-meter: account-level pace/quota guidance (#151) (4ea4dd1)
• cost-meter: attribute spend per command and per fix-loop iteration (#147) (05203d5)
• cost-meter: attribute spend per orchestration phase (#148) (522b428)
• session-review: /session-review skill + session-analysis agent (#154) (bccf6df)
• session-review: persist trend digest + harness-audit consumption (#155) (833c606)
• telemetry: capture agent-/auto-invoked skills distinctly; tighten bypass detection (#145) (2d025f1)

Bug Fixes

• extend prose-honesty gate to sibling docs and clean un-instrumented targets (#137) (e537495)
• pr: make /pr own the human gate for code review (#160) (e26175a)
• pr: make /pr own the human gate for code review (#165) (26741d5)

Documentation

• remove implemented and issue-converted design docs (#120) (11aa734)
• session-review: document OSS complements (ccusage, OpenTelemetry, claude-code-log) (#156) (71e3ed5)
• session-review: umbrella overview tying the harness together (#158) (ad3a21d)

6.4.0 (2026-06-06)

Features

• dev-team: add no-colon description rule to agent-audit, spec for commands→skills migration (c2a6a0d)
• dev-team: collapse commands/ into skills/ — unified capability layer (f9fde67)

Bug Fixes

• dev-team: update bats test paths and regenerate knowledge index after commands→skills migration (ca976e7)

6.3.0 (2026-06-05)

Features

• dev-team: /explore command, file-based /triage, agent-create reconcile (#56 #57 #58) (#94) (fb790d6)
• dev-team: automate test-layer-gates fixture as an agent-eval (#85) (#91) (84cfd35)
• dev-team: xUnit testing knowledge build-out (#73 #74 #75 #76) (#93) (60ca0ae)

6.2.0 (2026-06-05)

Features

• dev-team: add test-layer-gates knowledge file + fixture (#80) (884415d)
• dev-team: add test-strategy knowledge file (xUnit Test Strategy patterns) (0e7f58a)
• dev-team: add test-strategy knowledge file (xUnit Test Strategy patterns) (c101225)
• dev-team: behavior pre-gates + redundancy check for test-design-advisor (#80) (20b8053)
• dev-team: complete testing-strategy epic (#81 #82 #83 #84) (#90) (19abe6c)
• dev-team: register test-layer-gates + verify gate firings vs fixture (#80) (8d847cb)
• dev-team: skip /code-review for documentation-only changesets (f3aac6f)
• dev-team: skip /code-review for documentation-only changesets (2b75996)
• dev-team: wire behavior pre-gates + redundancy into test-design-advisor (#80) (f67b5a3)

Bug Fixes

• dev-team: repoint rule-id adapter contract refs + fix skill-wiring test (f7b0cb6)
• dev-team: repoint rule-id adapter contract refs + fix skill-wiring test (9f685a3), closes #65
• dev-team: sync /review alias frontmatter with /code-review (#88) (5ffd156)
• dev-team: sync /review alias frontmatter with /code-review (#88) (ae0ee28)

Documentation

• dev-team: document doc-only short-circuit in code-review-process (f59f008)

Miscellaneous

• convert pending specs/plans to GitHub issues; remove spec/plan files (aea265e)
• convert pending specs/plans to GitHub issues; remove spec/plan files (cf82c79)
• dev-team: resolve agent-audit compliance gaps (16b4aa9)
• dev-team: resolve agent-audit compliance gaps (ca89416)

6.1.0 (2026-06-04)

Features

• dev-team: add test-design and CD test-architecture capabilities (5281b37)
• dev-team: handle out-of-repo tests + document test evaluation workflow (42277cf)
• dev-team: outside-in baseline before refactor in test evaluation (7624645)
• dev-team: test design + CD test architecture capabilities (4177444)

6.0.0 (2026-06-02)

⚠ BREAKING CHANGES

• published plugin ids in the bfinster marketplace are now 'dev-team' and 'security-assessment' (previously 'agentic-dev-team' and 'agentic-security-assessment'). The 'agentic-' prefix carried no information — every plugin in this marketplace is agentic by definition.

Features

• upgrade: migrate legacy agentic-* plugin ids on upgrade (f6865fc)

Code Refactoring

• agents: orchestration cluster has no remaining sweep work (12c) (a7c3211)
• dev-team: sweep internal references to dev-team (5ce4ba8)
• rename plugins to dev-team and security-assessment (a36bba2)

Documentation

• readme: document /upgrade right after the install section (7411abf)
• repo-root: sweep references; add Renamed plugins README notice (49066fb)

5.6.0 (2026-06-02)

Features

• hooks: PostToolUse hook regenerates knowledge index + fail-open (568ed0a)
• hooks: pre-commit sibling hook + shared commit-detection helper (4db7eba)
• knowledge-index: pin jq >= 1.6 for stable output formatting (f66ceca)
• knowledge-index: ship the initial knowledge/index.json (4201181)
• knowledge-index: summary extraction with operational sentence boundary (cd4fe65)
• on-demand knowledge index + 550× perf rewrite + agent rename (4e680fc)

Code Refactoring

• agents: cite knowledge anchors in code-quality cluster (12b) (d2e50bc)
• agents: cite knowledge anchors in security cluster (12a) (9eb5a32)
• agents: orchestration cluster has no remaining sweep work (12c) (a7c3211)
• agents: rename files to match internal agent names (1b6d304)
• agents: rename refactoring-review → refactor-opportunity-review (d602e69)

Performance Improvements

• knowledge-index: rewrite builder inner loop as one Python process (1f3c06c)

Documentation

• capture knowledge indexing decision in ADR 0005; retire spec + plan (f1b291a)
• orchestrator: document the index lookup → section Read consumer pattern (f299182)

5.5.0 (2026-06-01)

Features

• commands: add /model-routing-check diagnostic (5aa05fc), closes #37
• environment-aware model routing with PreToolUse hook enforcement (511ec58)
• hooks: add codegraph-nudge skeleton with .codegraph/ presence check (f0fbcf2)
• hooks: codegraph-nudge blocks in careful mode (9c1ddf5)
• hooks: codegraph-nudge warns on Grep/Glob multi-file shape (fddaabd)
• hooks: PreToolUse Agent hook enforces pre-dispatch resolution (ff937a2), closes #37
• hooks: register codegraph-nudge and codegraph-turn-mark in settings.json (60f0fa0)
• hooks: sentinel-based turn-boundary detection for codegraph-nudge (9115852)
• init-dev-team: opt-in probe of /v1/models with three failure modes (d3fc9ec), closes #37
• init: bootstrap JS project via js-project-init when no package.json (8853172)
• init: state-aware CodeGraph step in /init-dev-team (10a62cf)
• knowledge: add adversarial review protocol, design smells, object calisthenics, testability patterns (d3fe547)
• knowledge: adversarial review protocol, design smells, object calisthenics, testability patterns (b75b8ec)
• model-resolve: happy-path tier→snapshot resolution (7affb2b), closes #37
• model-resolve: overrides, cascade, cycle, exhaustion, dump-map (3557378), closes #37
• model-resolve: perf gate + happy-path fast-path (069cfb6), closes #37
• model-routing: ship knowledge/model-routing.json defaults (e326cc1), closes #37
• skills: add adr-tools skill for npryce/adr-tools CLI mechanics (f0cce9d)
• skills: add mermaid-diagramming skill with blue-gray theme (f909895)
• state-aware CodeGraph integration for init flows + PreToolUse nudge hook (117a78e)
• ux: SessionStart hook surfaces routing overrides banner (9150c43), closes #37

Bug Fixes

• hooks: address inline-review findings (3f39271)
• move ADRs to docs/adr/ to match project convention (d568765)
• review: address code-review findings before PR (36f1b57)

Code Refactoring

• orchestrator: relocate model routing authority to PreToolUse hook (66bca9f), closes #37

Documentation

• adr: pre-dispatch model resolution + hook enforcement decisions (aa52c37), closes #37
• complete the hook-as-authority sweep + fix probe invocation path (0701731), closes #37
• complete the routing-doc cleanup + add architecture diagrams (2ab3725), closes #37
• document codegraph-nudge hook and updated init-dev-team flow (414f67b)
• fix CHANGELOG (0f7fd09)
• mention codegraph-nudge in agent-architecture and reference (3de9e94)
• model routing contract and troubleshooting guide (dc4bd03), closes #37

Miscellaneous

• remove pinned snapshot IDs outside routing.json (1d5f133), closes #37

5.4.0 (2026-05-28)

Features

• command: add /init-dev-team command and update advisory messages (62b9648)
• command: add Windows support to /init-dev-team (179874c)
• hook: add jq and python3 hard dependency guards (cbdb518)
• hook: blocking output, exit codes, and end-to-end JS/TS flow (534ae66)
• hook: language adapter dispatch with explicit adapter contract (3a9312b)
• hook: mutation-gate scaffold with fast-path, opt-out, and _timeout() (60b092e)
• hook: pitest Java adapter with runner-stdout test-list derivation (32b235f)
• hook: RED-GREEN transition detection with state file and stdout capture (92862a6)
• hook: register mutation-gate in PostToolUse Bash hook chain (a785606)
• hook: Stryker JS/TS adapter with fixture-based tests (53e3684)
• hook: Stryker.NET C# adapter (reuses parse_stryker_kills from lib) (c5221d4)

Bug Fixes

• hook: address spec-compliance review findings (baa072f)

Miscellaneous

• remove implemented plans and specs, add codegraph gitignore (41e64d8)

5.3.1 (2026-05-15)

Code Refactoring

• agent-skill-authoring: resolve overlap with agent-create skill (2818bd5)

5.3.0 (2026-05-14)

Features

• agent-create: add agent-create skill, official agent template, and schema validation (e872244)
• semantic-scan and agent-create skills with official schema validation (cc1b6b3)
• semantic-scan: add /semantic-scan skill and command for detecting logical duplication (324aea9)

Bug Fixes

• agent-create: move --dry check before file write; fix CLAUDE.md description (7f93ef3)

5.2.0 (2026-05-12)

Features

• add four missing subagent prompt templates (6c86def)
• add Skill tool to agents with ## Skills sections (c4eee0f)

Code Refactoring

• code-review: trim command file, move templates to output-format (6d04bbc)
• context-loading-protocol: drop stale token table, tighten (1972508)
• docker-image-create: tighten skill, keep runtime patterns inline (cd7e686)
• human-oversight-protocol: cut philosophy, tighten (8491232)
• js-project-init: collapse defaults into a list, drop rationale prose (a56257c)
• mutation-testing: drop overlap with constraints, trim (1e2ae4c)
• performance-benchmark: trim skill, move report template to examples (553a107)
• remove command wrappers and realign model routing (faf1cd8)
• specs: merge Constraints + Guidelines into one Rules list (ccff2d4)
• static-analysis-integration: extract maintenance, trim runtime skill (67ee544)
• tighten team agent prompts and add output discipline (0f36139)

Documentation

• add /explore spec, implementation plan, and exploratory-testing field guide (5f1ffec)
• add /triage file-based output spec and implementation plan (58b423f)

5.1.1 (2026-05-06)

Code Refactoring

• rename devops-sre-engineer to platform-engineer and fix doc drift (9d63904)

5.1.0 (2026-04-27)

Features

• security-assessment: ship apply-accepted-risks.sh + primitives contract v1.3.0 (caa62df)

5.0.0 (2026-04-24)

⚠ BREAKING CHANGES

• the /data-scientist slash command and data-scientist team agent have been removed. Consumers should migrate to /software-engineer or /architect.

Features

• recon: add optional file_inventory to envelope schemas (contract 1.2.0) (5dc9ffe)
• recon: canonical inventory enumeration script + ts-monorepo fixture (9bf2ded)
• remove data-scientist agent and overhaul plugin docs (a91d7e9)
• security-review: adapter error paths — malformed/unmapped category + missing category + bad mapping YAML (c70cbd1)
• security-review: adapter validates envelope schema + normalizes rule_id case + negative schema fixture (6a81033)
• security-review: agent output schema + judgment-only OWASP category annotations + reliability eval (095523d)
• security-review: canonical rule_id mapping + adapter happy-path (language-specific included) (02ce542)

Bug Fixes

• recon: align codebase-recon schema_version emission with 0.2 placeholder bump (6558664)

Code Refactoring

• security-review: strip pattern-visible classes from owasp-detection with pointer stubs (Item 3b) (9af6355)

Documentation

• agentic-dev-team: update cross-references to renamed companion plugin + history note on rename docs (87a7a34)
• overlap-cleanup: trigger-context section on security-review agent + reciprocal companion README note (e6f5378)
• recon: contract 1.2.0 + codebase-recon Step 6.5 + fail-open consumer contract + pipeline budget (af61d67)
• regenerate team-agents diagram for current roster (1016411)
• security-review: adapter docs + Phase 1b wiring + AST invariant + runtime smoke + backward-compat (0522025)
• specs: approved specs + plans for Item 5, Gap 6a, and plugin-rename (764aa3b)

4.0.0 (2026-04-22)

⚠ BREAKING CHANGES

• skill file paths changed from skills/foo.md to skills/foo/SKILL.md. Team agent count reduced from 12 to 11.

Features

• add baked-in config, Swiss Army Knife, and stateful container checks to docker-image-audit (eefe5a8)
• add codebase-recon agent with git history overview (577cf98)
• add default permissions to auto-approve most tools (440407c)
• add docker-image-create and docker-image-audit skills (7e115c8)
• add feature-file-validation skill to test-review pipeline (5f53264)
• add plan review personas, performance benchmarking, review-fix loop, and auto-scope (682e7ec)
• add static analysis pipeline integration to code-review (c031b4f)
• auto-trigger plan after spec approval and add BDD scenario review (af99078)
• bump primitives contract to v1.1.0 + lift reference implementation details into plan (edc02da)
• custom SARIF-emitting scripts — entropy-check + model-hash-verify (b15762e)
• guard primitives-contract edits with semver-bump requirement (730ccd1)
• js-project-init: add Husky pre-push hook and drop eslint-plugin-prettier (119a71a)
• namespace plugin as agentic-dev-team@bfinster (0a86eef)
• persist /specs output to docs/specs/ after consistency gate passes (69004a9)
• publish versioned security-primitives-contract v1.0.0 (eed5bf5)
• restructure skills into directories with progressive disclosure (bab081b)
• SARIF-first tool orchestration baseline (required 5 adapters) (f5ed4fe)
• support ACCEPTED-RISKS.md project-local policy carveouts (c40a16f)

Bug Fixes

• scope: CI/CD workflow files explicitly in scope for static + security review (763924f)
• update skill file references to include SKILL.md path (90cd81d)
• update skill file references to include SKILL.md path (34a7474)
• use official claude plugin update mechanism in /upgrade command (ab4c5d7)

Code Refactoring

• build: remove canned summary template; trust native progress output (8c2eb47)
• move hook registrations to plugin settings.json (95af67d)
• move plugin components into plugins/agentic-dev-team/ (b1a4792)
• split CLAUDE.md into plugin config and dev instructions (8157142)

Documentation

• move per-plugin install instructions into each plugin's README (26bca28)

Miscellaneous

• main: release 2.1.1 (66b5ad9)
• main: release 2.1.1 (118ed58)
• main: release 2.2.0 (ab37326)
• main: release 2.2.0 (a35dad7)
• main: release 2.3.0 (7b5ebe7)
• main: release 2.3.0 (182a222)
• main: release 3.0.0 (6825078)
• main: release 3.0.0 (8950366)
• main: release 3.1.0 (8413514)
• main: release 3.1.0 (36d22b6)
• main: release 3.1.1 (2e7db64)
• main: release 3.1.1 (4425e5d)
• main: release 3.2.0 (cfbdd77)
• main: release 3.2.0 (6165507)
• main: release 3.3.0 (766c71c)
• main: release 3.3.0 (c7b4597)

---

Pre-Restructure History (v2.0.0–v3.3.0)

These entries are from the root CHANGELOG.md that existed before the repository was restructured into a multi-plugin monorepo. Version tags in this section use the old vX.Y.Z format rather than the current agentic-dev-team-vX.Y.Z component-scoped format.

3.3.0 (2026-04-14)

Features

• add default permissions to auto-approve most tools (440407c)

Bug Fixes

• update skill file references to include SKILL.md path (90cd81d)
• update skill file references to include SKILL.md path (34a7474)

3.2.0 (2026-04-10)

Features

• add plan review personas, performance benchmarking, review-fix loop, and auto-scope (682e7ec)
• replace Mermaid diagrams with styled SVG images (0e11eb8)

Bug Fixes

• add more bottom padding to three-phase workflow SVG (151cfef)
• align fail-loop arrow to enter TDD box from the left edge (37c4aa8)
• clean up broken lines and misaligned arrows in three-phase SVG (86f97e9)
• compress three-phase workflow SVG to prevent GitHub clipping (2032cc9)
• connect Gate 3 arrows directly to /pr and Learning Loop box tops (99d2d32)
• increase three-phase workflow SVG viewBox height to prevent clipping (18441fd)
• move human gates inline with last step of each phase (2134bc5)
• replace T-junction with two direct lines to /pr and Learning Loop (703515b)

3.1.1 (2026-04-10)

Bug Fixes

• use official claude plugin update mechanism in /upgrade command (ab4c5d7)

3.1.0 (2026-04-10)

Features

• auto-trigger plan after spec approval and add BDD scenario review (af99078)

3.0.0 (2026-04-09)

⚠ BREAKING CHANGES

• skill file paths changed from skills/foo.md to skills/foo/SKILL.md. Team agent count reduced from 12 to 11.

Features

• add baked-in config, Swiss Army Knife, and stateful container checks to docker-image-audit (eefe5a8)
• add docker-image-create and docker-image-audit skills (7e115c8)
• restructure skills into directories with progressive disclosure (bab081b)

2.3.0 (2026-04-08)

Features

• add feature-file-validation skill to test-review pipeline (5f53264)
• namespace plugin as agentic-dev-team@bfinster (0a86eef)
• persist /specs output to docs/specs/ after consistency gate passes (69004a9)

2.2.0 (2026-04-06)

Features

• add static analysis pipeline integration to code-review (c031b4f)
• js-project-init: add Husky pre-push hook and drop eslint-plugin-prettier (119a71a)

2.1.1 (2026-04-02)

Code Refactoring

• move hook registrations to plugin settings.json (95af67d)
• move plugin components into plugins/agentic-dev-team/ (b1a4792)
• point marketplace.json source to plugins/agentic-dev-team (b5ee9b8)
• split CLAUDE.md into plugin config and dev instructions (8157142)

2.1.0 (2026-04-02)

Features

• add /version command to report installed plugin version (712ee1e)

2.0.0 (2026-04-02)

⚠ BREAKING CHANGES

• The /beads command and beads skill are no longer available. Users relying on bd for task tracking should use memory/ progress files and /continue instead.

Features

• add js-project-init skill for scaffolding JS projects (e753742)
• automated pre-commit code review gate (67df646)
• remove beads task tracking from plugin (fd2444f)

Bug Fixes

• prevent false positive on gitignored .env files in security-review (cb2b002)

Code Refactoring

• mutation testing skill to use real tools instead of academic estimation (892168b)