This page documents every user-invocable command in the dev-team plugin by capability group.
Commands within each group are split into two tiers:
Multi-agent (orchestrators) — dispatch more than one Agent tool call per invocation, directly
or via delegation, and may include human gates.
Standalone (single-pass) — single-agent or self-contained; return in one pass without
inter-phase human gates.
The two multi-phase pipelines with inter-phase gates are /ship —
whose full phase table appears at the end of this page — and
/test-improve, whose phase reference lives on its own page,
test-improve.md.
Execute an approved plan in small per-behavior batches (Code-First Small Batches) with inline review checkpoints and verification evidence
[--plan <path>] [--yes]
/setup
skills/setup/SKILL.md
Provision a repo end to end: install prerequisites, generate project config, activate agent templates
[--yes] [--dry-run]
/project-init
skills/project-init/SKILL.md
Detect the stack, inventory tools, install only what's missing; offers opt-in graph-tools
[--yes]
/autoship
skills/autoship/SKILL.md
Orchestrate a bounded round of automated issue processing: reclaim orphaned in-progress issues, discover autoship:ready issues, and invoke /ship for each until a cost or count cap is hit
--max-issues N --max-cost-usd N [--dry-run] [--label LABEL]
Evaluate an app's tests and recommend a CD-pipeline-aligned test architecture — fast, deterministic tests that validate behavior in CI without configuring the rest of the system
Convergence worker: iterate toward the four quality targets (coverage ≥ 90%, zero surviving mutants, determinism, fastest pre-merge wall-clock), dispatching the smallest action that closes the largest gap each round
Convert a /cd-test-architecture assessment into a parent + Phase-tagged child issues on the operator's tracker (ADO, GitHub, GitLab, Jira); falls back to local plan files when the CLI is missing
Build a xunit.v2 Stryker shim so Stryker.NET produces a valid mutation score for a xunit.v3 test project; run it before mutation testing a .NET project on xunit.v3
Compare this plugin against external plugins, tools, or feature sets
<comparison target>
/context-loading-protocol
skills/context-loading-protocol/SKILL.md
Decide which agents and skills to load for a given task
[task description]
/feedback-learning
skills/feedback-learning/SKILL.md
Capture amend/learn/remember/forget keywords and update agent or skill behavior
<feedback statement>
/handoff
skills/handoff/SKILL.md
Compress or split off context for another session to pick up
[--compress \| --split]
/human-oversight-protocol
skills/human-oversight-protocol/SKILL.md
Clarify approval gates, intervention commands, and transparency requirements
[context]
/review-summary
skills/review-summary/SKILL.md
Generate compact session summary for context continuity
[--from <json-file>]
/long-eval
skills/long-eval/SKILL.md
Run an eval that outlives a single cloud-session container — agent calibration, prompt sweeps, judge-panel scoring — so it survives container recycles and can be resumed
Report actual token spend and dollar cost of dispatched work and flag regressions
[--transcript <path>] [--tolerance <n>]
/artifact-lifecycle
skills/artifact-lifecycle/SKILL.md
Report on skill and agent usage data from ~/.claude/metrics/artifact-usage.json
[--json]
/telemetry
skills/telemetry/SKILL.md
Manage and report the opt-in usage telemetry beacon
[on\|off\|status\|report]
/version
skills/version/SKILL.md
Report the installed plugin version
(none)
/upgrade
skills/upgrade/SKILL.md
Check for and apply plugin updates from within a session
(none)
/help
skills/help/SKILL.md
List the main dev-team workflows; --all shows every user command
[--all]
/agent-readiness
skills/agent-readiness/SKILL.md
Score how ready the current repo is for AI-assisted development against the Agent-Readiness Scorecard and emit a tiered report
[repo-path] [--json <file>] [--markdown <file>]
/run-report
skills/run-report/SKILL.md
Report one orchestrated run's timeline — per-state dwell time, rejection count, hook denials/bypasses, and cost — joined from the boundary/cost/state event logs
[--session <id>]
/report-pdf
skills/report-pdf/SKILL.md
Render a dev-team Markdown report (.dev-team-reports/ or reports/) to a polished, shareable PDF
File:skills/ship/SKILL.mdRole: orchestrator.
Use when: the user says "ship this", "take this feature end to end", or
wants the spec → plan → build → PR flow without re-assembling it each time.
Plan — decompose into vertical slices with Gherkin scenarios; a tier-scaled set of plan-review personas (1–5, by plan complexity) runs in parallel before the gate.
Build — small per-behavior batches per slice (Code-First Small Batches), inline review checkpoints, verification evidence. Do not proceed until the suite is green.
Report — PR URL, quality-gate result, whether auto-merge is armed.
(orchestrator only)
—
Agents involved (dispatched by the delegated skills)¶
/build's inline review checkpoints dispatch the review agents listed in
team-structure.md → Review Agent Dispatch.
/plan dispatches a tier-scaled subset of the five plan-review persona
agents (agents/plan-review-*.md) — the Acceptance Test Critic always
runs; the rest are added as the plan's tier (trivial/standard/complex)
warrants — and the
progress-guardian gate-keeper.
/code-review re-runs the same review agents over the full changeset.
Use /test-improve alongside /ship for multi-phase quality pipelines — run
/test-improve to raise baseline coverage and health first, then /ship to
carry the feature through spec → build → PR.
/ship and /test-improve are the two multi-phase pipelines with
inter-phase human gates in the plugin. Every other slash
command is either a single-step worker (e.g. /coverage-baseline,
/triage) or a one-shot orchestrator that returns in a single pass (e.g.
/code-review, /test-design). Knowing the phase order, the owning skill
or agent for each step, and where the human gates fall is the difference
between operating these workflows confidently and re-reading every SKILL.md
each time.
One cross-command lifecycle is documented separately: the defect workflow
that connects /triage's triage records, /code-review's correction
prompts, and /apply-fixes — from discovery to applied fix, including who
owns leftover corrections/ files — lives in
triage-workflow.md.