메뉴
Review branch changes or a PR for code quality, tests, documentation, and commit hygiene. Routes to `reviewer` for own work or `pr-reviewer` for a cross-author PR. Invoke with /review-changes.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Generates, stress-tests, and iteratively evolves ideas for a stated problem — product concepts, features, solution options, strategies — using research-grounded divergent/convergent agent loops: parallel persona generators (nominal-group simulation), independent judges scoring novelty, feasibility, impact, and fit on separate axes, and bounded recombination rounds gated by /confidence. Auto-triages run depth (quick in-context vs deep multi-agent; override with quick|deep). Use when brainstorming, exploring solution options, or pressure-testing a concept. Triggers on "brainstorm", "give me ideas", "help me come up with", "ideate on", "what could we build", "/ideate".
Reviews whether a change takes the most optimal approach for its stated intent and, when it does not, researches the code, validates a concretely better approach via holistic analysis, and either proposes it (report mode) or applies it behind a confidence gate (apply mode). Judges four axes — codebase-fit, simplicity, performance, robustness — at the approach level, deferring line-level and failure-mode findings to code-quality, critical, and holistic-review. Stays silent when the approach is already optimal (quiet early-exit). A `plan` mode reviews a drafted plan's approach at plan time (aw-planner Phase 1) — the cheapest moment to switch. Called by the reviewer and pr-reviewer agents, the polish skill, and aw-planner as a default-on lens; also runnable standalone. Triggers on "is this the best approach", "better way to do this", "is this optimal", "optimize this approach", "rethink the approach", "/optimize-approach".
Execute complete feature development cycles autonomously — from task intake through tested PR delivery — using isolated Git worktrees. Phase-based workflow (0–7) with optional companion skills for planning, quality gates, TDD, UX, code quality, docs, and CI verification. Companions skip silently if not installed. Triggers on "implement autonomously", "end-to-end", "in isolation", "in a worktree", or independent feature work. Invoke with /autonomous-workflow.
Create the plan artifact (plan.md + checks.yaml executable acceptance checks) in `.agent/{branch}/` from the current conversation context. Captures all Phase 0-1 discussion into a structured, self-contained document that enables context recovery and session handoff. On every invocation, writes plan.md and re-derives checks.yaml from the Acceptance Criteria; an immutable plan.vN.md snapshot is written only in opt-in snapshot mode (`Skill("aw-create-plan", "snapshot")`). Use after planning is complete and confidence gate passes — and again on every plan iteration (user-requested refinement or Phase 4 auto-replan). Triggers on create plan, generate plan, write plan artifact, regenerate plan, iterate on plan.
Resolves a single bug from any starting evidence — Dash0 telemetry (span / log / web event / RUM error link), raw stack trace, error message, code pointer (file:line), screen recording, Linear ticket URL, or free-text symptom. Classifies the input, **triages complexity** (Phase 0.5) to pick between a fast lane and a full holistic-analysis lane, runs a pre-flight sweep, locks a failing reproduction (delegating to /tdd, /e2e-testing, or /e2e-testing-mobile by layer), delegates root-cause analysis to the isolated rca-investigator agent (holistic-analysis + confidence) on complex bugs (or runs a lightweight in-skill analysis on simple ones), gates on confidence(analysis), and on >= 92 % hands off **without human confirmation**: simple bugs take the fast lane (/fix-bug → aw-create-plan → aw-executor, no aw-planner) and complex bugs take the standard lane (aw-planner → aw-executor), both with a CEGIS refinement contract. Fast-lane round-3 CEGIS failure falls back to standard-lane via aw-planner. An independent bug-
Forces a full holistic re-analysis when a fix or refactor isn't working. Instead of continuing to patch in isolation, this skill triggers a structured step-back analysis that traces the entire execution path end-to-end — from entry point to exit — analyzing each block, every contract boundary, and the full data flow. Three modes: "fix" (default) for bugs and broken behavior, "refactor" for restructuring/improvement, and "review" for PR validation (returns structured intent-match + system-fit findings for the reviewer / pr-reviewer agents to consume — never run on its own for routine review work). Trigger ONLY when at least one isolated fix attempt has already failed, or the user explicitly requests a full execution-path analysis — phrases like "step back", "think holistically", "analyze the whole thing", "zoom out", "look at the bigger picture", or "rethink this" qualify only in that context. Never trigger for trivial one-line fixes or first-attempt debugging. Also triggers on "/holistic", "/step-back", "/ret
| name | review-changes |
| description | Review branch changes or a PR for code quality, tests, documentation, and commit hygiene. Routes to `reviewer` for own work or `pr-reviewer` for a cross-author PR. Invoke with /review-changes. |
| disable-model-invocation | true |
| argument-hint | [<pr-url>|#<n>] [--report] [--critical] [--publish] |
| license | MIT |
| metadata | {"author":"mthines","version":"2.0.0","workflow_type":"command"} |
Choose the agent based on the argument shape:
| Argument shape | Agent | Reason |
|---|---|---|
no arg / --report | reviewer | Own branch (with or without an own PR). reviewer auto-detects Fix / Report / Self-Review. |
PR URL or #<n> | run gh pr view --json author first; if author == current user → reviewer Self-Review, else → pr-reviewer | |
--publish | always pr-reviewer (--publish is the cross-review authorization token) |
# Default — own work
Agent(subagent_type: "reviewer", prompt: "Review changes. Arguments: $ARGUMENTS")
# Cross-review (PR URL or #n with cross-author, OR --publish present)
Agent(subagent_type: "pr-reviewer", prompt: "Review PR. Arguments: $ARGUMENTS")
| Invocation | Effect |
|---|---|
/review-changes | Fix Mode on the current branch — auto-fix simple, plan complex. |
/review-changes --report | Report Mode — propose only, no auto-fixes. |
/review-changes --critical | Adds adversarial pre-mortem (Skill("critical", "code")). |
/review-changes --with a,b,c | Loads up to 3 additional review lenses. |
/review-changes <PR-URL> | Routes to pr-reviewer if cross-author; to reviewer Self-Review if own PR. |
/review-changes <PR-URL> --publish | Cross-review with authorization to post as a pending GitHub review. |
--commentsThe old --comments flag is gone. Cross-review with line-level inline comments now lives in the pr-reviewer agent and is the default behaviour when a cross-author PR is passed. Authorization to post a pending review is granted via --publish (token path) or an explicit authorization phrase in the chat ("publish them", "post them", "go ahead and post", "submit the review") — see agents/pr-reviewer/rules/authorization-gate.md.
Without authorization, pr-reviewer produces the comment proposal in the terminal and stops. The user reads the proposal and decides whether to re-invoke with --publish or paste comments manually.