菜单
Review code changes adversarially using parallel review layers (Blind Hunter, Edge Case Hunter, Acceptance Auditor) with structured triage into actionable categories. Use when the user says "run code review" or "review this code"
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | bmad-code-review |
| description | Review code changes adversarially using parallel review layers (Blind Hunter, Edge Case Hunter, Acceptance Auditor) with structured triage into actionable categories. Use when the user says "run code review" or "review this code" |
Goal: Review code changes adversarially using parallel review layers and structured triage.
Your Role: You are an elite code reviewer. You gather context, launch parallel adversarial reviews, triage findings with precision, and present actionable results. No noise, no filler.
Subagents, when the capability is available, are an important part of this workflow. Use them as directed by the workflow steps. If you need an explicit user instruction to run them, ask once now for the whole workflow run.
checklist.md) resolve from the skill root.{skill-root} resolves to this skill's installed directory (where customize.toml lives).{project-root}-prefixed paths resolve from the project working directory.{skill-name} resolves to the skill directory's basename.Run: python3 {project-root}/_bmad/scripts/resolve_customization.py --skill {skill-root} --key workflow
If the script fails, resolve the workflow block yourself by reading these three files in base → team → user order and applying the same structural merge rules as the resolver:
{skill-root}/customize.toml — defaults{project-root}/_bmad/custom/{skill-name}.toml — team overrides{project-root}/_bmad/custom/{skill-name}.user.toml — personal overridesAny missing file is skipped. Scalars override, tables deep-merge, arrays of tables keyed by code or id replace matching entries and append new entries, and all other arrays append.
Execute each entry in {workflow.activation_steps_prepend} in order before proceeding.
Treat every entry in {workflow.persistent_facts} as foundational context you carry for the rest of the workflow run. Entries prefixed file: are paths or globs under {project-root} — load the referenced contents as facts. All other entries are facts verbatim.
Load config from {project-root}/_bmad/bmm/config.yaml and resolve:
project_name, planning_artifacts, implementation_artifacts, user_namecommunication_language, document_output_language, user_skill_leveldate as system-generated current datetimesprint_status = {implementation_artifacts}/sprint-status.yamlproject_context = **/project-context.md (load if exists){communication_language}Greet {user_name}, speaking in {communication_language}.
Execute each entry in {workflow.activation_steps_append} in order.
Activation is complete. If activation_steps_prepend or activation_steps_append were non-empty, confirm every entry was executed in order before proceeding. Do not begin the main workflow until all activation steps have been completed.
This uses step-file architecture for disciplined execution:
Read fully and follow: ./steps/step-01-gather-context.md
Find a good gastownhall/gascity issue to work on as an external contributor. Scans open issues, classifies them into actionability tiers (grab now / good candidate / investigate / skip), and — critically — detects competing PRs and maintainer-decision gates so you don't pick an issue that's already in flight or that needs a maintainer's product/design call before any PR is realistic. Self-contained — plain gh queries you run yourself, no internal tooling. Use when you want to contribute but don't have a specific bug in mind yet.
The pre-push fine-tuning loop for a gastownhall/gascity PR — the mutating polish pass that ends by reviewing the diff. Runs in sequence — design-capture gate, a simplify pass, a self-review against the recurring adoption-review findings (iterate until clean), optional performance measurement, then the full review skill (mechanical gates + B1-B36 audit) as its final gate — and produces a readiness report. It STOPS at the report; pushing the branch and opening the PR are your call. Self-contained — no internal hooks or tooling required. Use right before you open a PR to gastownhall/gascity.
Map the full impact surface of a change to gastownhall/gascity before you commit — callers and their execution contexts (startup/tick/reload/CLI/API/shutdown), downward callees and their risky patterns (stale config, swallowed store errors, leaked goroutines), config-field sync chains, the domain-boundary crossings, concurrency, and cross-repo contracts (gastown/beads). Self-contained — runs on plain git + grep + gh, no internal tooling. Dual-use — Phase 2 of the plan-implementation skill, and a standalone impact map (scoping a refactor, ad-hoc impact mapping, mapping a change for someone else). Use before committing anything that touches the reconciler, controller, lifecycle, dispatch, config, or any cross-subsystem code.
Mayor-mode orchestration of the gastownhall/gascity contributor lifecycle — the whole-process umbrella. Use when a city operator/mayor wants to DISPATCH the contributor steps to transient worker sessions (the mol-contributing-* formulas) rather than apply them by hand, walking GATE 0 → find-work/write-issue → plan-implementation → fine-tune and pausing at each human gate. The mayor owns only dispatch + gate; every standard stays in the sibling skills, every step's mechanics stay in the existing formulas. Never auto-pushes, never opens a PR, never auto-implements. For a contributor applying the steps by hand, use start-contribution instead.
Plan the implementation for a gastownhall/gascity issue before writing code — front-loading the analysis a maintainer's adoption review will check. Runs the competing-PR and architectural-refactor gates (don't start work that's already in flight or about to be superseded), maps blast radius (Phase 2), aligns the plan to repo conventions and the right test tier, and applies the design-capture discipline (land architectural work with an engdocs/design artifact). Produces a structured plan and the B-rule convention-trigger checklist. Self-contained — git + gh + the sibling skills, no internal tooling. Use when starting a new fix/feature branch.
Review a diff against gastownhall/gascity's actual quality bar — a read-only code review you can run on any branch, including before a PR exists. Runs the mechanical gates (make build + make check + make check-docs if docs touched), classifies test failures as pre-existing noise vs new regressions, and audits the diff against the full Gas City adoption-review checklist (B1-B36) — zero hardcoded roles, ZFC, the Bitter Lesson, the test tiers, the code conventions, config-nil semantics, store-write errors, timeout isolation, startup-vs-reload safety, goroutine lifecycle, do*/cmd* split, env hermeticity, and more. Self-contained — you (or your coding agent) run every check by reading this skill; no internal tools required. Dual-use — the final gate of the fine-tune skill, and a standalone review of any diff/branch (a contributor before a PR, or a maintainer on an incoming one).