Agent skill

Harness Review Dispatcher

by Chachamaru127 in Chachamaru127/claude-code-harness

Routes a review request to the right mode and reference file in the claude-code-harness project: code, plan or scope review, a quick closeout, a dual Claude-and-Codex opinion, or a security-only pass.

MITAuto-check: notesDevelopment

SKILL.md written in Japanese; this summary is our English description.

Install Harness Review Dispatcher

skills CLI
$ npx skills add Chachamaru127/claude-code-harness --skill harness-review -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Chachamaru127/claude-code-harness harness-review --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Chachamaru127/claude-code-harness.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/harness-review .claude/skills/harness-review && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
harness-review
GitHub stars
3.2k
Token cost
~3.6k tokens
SKILL.md length
1,079 words
Files
13 (incl. references)
Skills in repo
25
Repo updated
First seen
Licence
MIT

At a glance

Routes a review request to the right mode and reference file in the claude-code-harness project: code, plan or scope review, a quick closeout, a dual Claude-and-Codex opinion, or a security-only pass.

  • Works in 3 steps: working tree: staged / unstaged /… → branch range: upstream または main/master… → recent commits: clean tree で branch…
  • Reviewing the code just written for quality and security issues
  • SKILL.md covers Dispatcher Contract, Quick Reference, Mode Decision and Pre-Review Cursor…, plus 12 more sections
  • Calls bash and claude

What it does

This Japanese-language skill is a thin dispatcher for a multi-angle review harness: the detailed quality criteria live in the references folder, not in the main file. With no argument it interprets the request as reviewing the work done so far, auto-detects the review target, and either starts automatically when one target is clear or asks the user to choose when it is ambiguous or there are several candidates.

A quick-reference table maps commands to modes: plain code review, a quick closeout for small dirty changes, a Codex-advised closeout, a dual Claude-and-Codex second opinion, a Cursor pre-review or second opinion, a forced team debate, a security-only review, and dedicated plan and scope reviews, each loading only the reference files its mode needs, such as code-review.md, security-profile.md or scope-review.md.

The skill's own contract is read-only: by default it never commits, pushes or releases, even on an APPROVE verdict, and delegates any actual commit to harness-work or harness-release unless the user explicitly asks or an opt-in flag is designed for it. When output is relayed through a local command stdout channel, the last line must carry a fixed literal telling the user that Claude will summarize the result and that Enter or a new prompt continues, so the review doesn't look like it stalled.

When your agent uses it

  • Reviewing the code just written for quality and security issues
  • Reviewing a plan file before implementation starts
  • Checking a change for scope creep or missed requirements
  • Getting a second opinion from Codex or Cursor alongside the primary review

Example prompts

  • “/harness-review 今までの作業をレビューして。”
  • “/harness-review --security このdiffのセキュリティレビューをして。”
  • “/harness-review plan Plans.mdの計画をレビューして。”

Requirements

  • The claude-code-harness project
  • Pre-approved tools (allowed-tools): Read, Grep, Glob, Bash, Task, Monitor, AskUserQuestion

Workflow steps

3 steps, taken from the first numbered list in SKILL.md.

  1. working tree: staged / unstaged / untracked を含む未コミット変更のみ
  2. branch range: upstream または main/master から HEAD までの commits
  3. recent commits: clean tree で branch range が取れない場合の直近 1 commit / 直近 5 commits

What it can do on your machine

Read from SKILL.md and the folder at commit 2b2b748. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves these tools, so the agent can use them without asking each time:

    • Read
    • Grep
    • Glob
    • Bash
    • Task
    • Monitor
    • AskUserQuestion

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • bash
    • claude

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Harness Review Dispatcher loads about 3.6k tokens when it runs, and up to ~21k if it reads all its reference files. Until then it costs about 53 tokens; SKILL.md has 1,079 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~53
When it runs · the whole SKILL.md, loaded when a task matches
~3.6k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~21k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check: notes

The automated check noted patterns worth knowing about, such as sudo or a known installer.

  • NotePre-approves every shell command (allowed-tools: Bash)SKILL.md
    allowed-tools: Read, Grep, Glob, Bash, Task, Monitor, AskUserQuestion

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Chachamaru127/claude-code-harness at commit 2b2b748, republished under its MIT licence (© Chachamaru127). 1,079 words, ~3,622 tokens.

Download SKILL.mdSave it as .claude/skills/harness-review/SKILL.md (or your agent's skills folder). This skill also uses 12 other files; get the full folder from GitHub.
name
harness-review
description
HAR: Multi-angle code, plan, scope review. Security/quality check. Trigger: review, code review, plan review, scope analysis. Do NOT load for: implementation, new features, bugfix, setup, release.
allowed-tools
Read, Grep, Glob, Bash, Task, Monitor, AskUserQuestion
description-en
HAR: Multi-angle code, plan, scope review. Security/quality check. Trigger: review, code review, plan review, scope analysis. Do NOT load for: implementation…
description-ja
HAR:コード・プラン・スコープを多角的にレビュー。セキュリティ・品質チェック。レビューして、レビュー、コードレビュー、プランレビュー、スコープ分析で起動。実装・新機能・バグ修正・セットアップ・リリースには使わない。
kind
workflow
purpose
Review code, plans, scope, and evidence before acceptance
trigger
review, レビューして, code review, plan review, scope analysis
shape
evaluate
role
evaluator
pair
harness-work
owner
harness-core
since
2026-05-05

Harness Review

Harness の統合レビュースキル。 この SKILL.md は薄い dispatcher であり、詳細な品質基準は references/ を読む。

if $ARGUMENTS == "": → 「今までの作業のレビュー」と解釈し、Review target detection を実行する → review target が 1 つに確定できる場合だけ自動開始する → review target が不明または複数候補の場合は AskUserQuestion で選択肢を出し、認識を揃えてから開始する

<!-- 上記 3 行は AUTO-START CONTRACT。skill-editing.md の「最冒頭 3 行以内」ルールに従い fence / HTML コメントで押し下げない -->
Output Contract (P35: 「止まったように見える」UX 対策)

<local-command-stdout> で host へ結果を中継する場合だけ、output の 最後の 1 行に次の literal を含める:

↑この結果は Claude が要約します。Enter キーで次へ進むか、新規 prompt で別の指示を出してください。

これは <local-command-stdout> 経由で text response として表示されると user が「止まった」と感じる UX 問題への明示的な instruction (patterns.md P35)。 host の最終回答は verdict、判断理由、確認した証拠、未検証点を返す。要約の予告だけで終了しない。

Dispatcher Contract

この skill の責務は review 判定だけ。 commit / push / release は既定では行わない。

  • review default read-only boundary: 既定は read-only。APPROVE でも自動 commit しない
  • Do not push just to review: review 目的だけで push しない
  • commit が必要な場合は、ユーザー明示依頼、harness-work、または harness-release の Work Commit Gate に委譲する
  • --commit-on-approve のような明示 opt-in が設計されるまで、この skill 単体の default side effect は禁止

Quick Reference

CommandModePurpose
/harness-reviewcode今までの作業を自動検出して review
/harness-review --quickquick小さな dirty change を軽く closeout
/harness-review --codex-closeoutcodex-closeoutCodex 助言 + focused tests で closeout
/harness-review --dualdualClaude + Codex second opinion
--pre-review cursorcode+pre-reviewfresh-context composer advisory pre-review before brain verdict(read-only)
/harness-review --cursorcode+cursor-second-opinioncore review gates + cursor second-opinion(brain 一次レビュー必須)
HARNESS_IMPL_BACKEND=cursor harness-reviewcode+cursor-second-opiniondefault ON 時も core review gates に cursor second-opinion を自動加算。primary verdict は brain 固定
/harness-review --team-debateteam-debateTeamAgent Debate を強制
/harness-review --securitysecuritysecurity 専用 review
/harness-review planplanPlans.md の計画 review
/harness-review scopescopescope creep / 漏れ review

Mode Decision

引数から実行 mode を決定し、必要な references/ を選択ロードする。

入力mode読む reference
引数なし / codecodereferences/code-review.md, references/governance.md
--quickquickreferences/codex-closeout.md, references/code-review.md
--codex-closeoutcodex-closeoutreferences/codex-closeout.md
--dualdualreferences/dual-review.md, references/team-debate.md
--team-debateteam-debatereferences/team-debate.md, references/governance.md
--securitysecurityreferences/security-profile.md, references/governance.md
--ui-rubricui-rubricreferences/ui-rubric.md
planplanreferences/plan-review.md, references/governance.md
scopescopereferences/scope-review.md, references/governance.md
--cursor or resolver result cursor for no-arg / code review onlycode+cursor-second-opinionreferences/code-review.md, references/governance.md, references/cursor-review.md, references/dual-review.md
fullfullreferences/code-review.md, references/team-debate.md, references/dual-review.md

quick と codex-closeout は軽量 path。 小さな dirty change、single commit、PR branch の closeout を速く見る。 品質 gate を捨てるものではない。

Cursor Default ON

mode 判定時、明示 mode words (plan, scope, full) と明示フラグを先に確定する。no-arg / code review の場合だけ helper root を解決して resolver を 1 回だけ実行し、resolver 不在時は claude とみなす。

bash
HARNESS_PLUGIN_ROOT="${HARNESS_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT:-}}"; if [ -z "$HARNESS_PLUGIN_ROOT" ] && [ -n "${CLAUDE_SKILL_DIR:-}" ]; then probe="$(cd "${CLAUDE_SKILL_DIR}" && pwd)"; while [ "$probe" != "/" ] && [ ! -d "$probe/scripts" ]; do probe="$(cd "$probe/.." && pwd)"; done; [ -d "$probe/scripts" ] && HARNESS_PLUGIN_ROOT="$probe"; fi
if [ -x "${HARNESS_PLUGIN_ROOT:-}/scripts/resolve-impl-backend.sh" ]; then resolved_backend="$(bash "${HARNESS_PLUGIN_ROOT}/scripts/resolve-impl-backend.sh" --role reviewer)"; else resolved_backend="claude"; fi

no-arg / code review で結果が cursor の場合は --cursor と同じ cursor-second-opinion を追加するが、core review gates (references/code-review.md, references/governance.md) は必ず先に読み、Cursor reference は additive にだけ扱う。primary verdict は brain 側で維持し、cursor は dual_review.cursor_verdict の advisory に限る。plan / scope など明示 mode word は resolver result より優先し、cursor default によって plan/scope references も code/governance references も置き換えない。結果が claude / codex の場合は従来どおりで、review の primary 判定面は変えない。

Pre-Review Cursor (--pre-review cursor)

/harness-review --pre-review cursor は bash scripts/pre-review-cursor.sh [--base ref] で read-only fresh-context composer pre-review(model-routing.sh --host cursor --tier review → cursor-companion.sh task、--write / --workspace / --resume なし)を 1 回実行し、PRE_REVIEW_FINDINGS: を brain 一次レビュー入力へ添付。companion 失敗は PRE_REVIEW_SKIPPED + exit 0(fail-open)。verdict は brain のみ(self-review scope 契約)。

Review Target Detection

REVIEW_AUTOSTART 契約: 引数なし ($ARGUMENTS == "") で呼ばれた場合、review / /review / /harness-review だけの入力を「今までの作業のレビュー」と解釈する。 Step 1 開始前の handshake 行として次を 1 行だけ出力する。

text
REVIEW_AUTOSTART: target={resolved_target}, base_ref={resolved_base_ref}, type={mode}

REVIEW_TARGET_ASK 契約: bare 呼び出しで review target が不明または複数候補の場合、Step 1 に進む前に AskUserQuestion を 1 回だけ使い、候補を 2-3 個に絞って確認する。

候補は次の順で作る。

  1. working tree: staged / unstaged / untracked を含む未コミット変更のみ
  2. branch range: upstream または main/master から HEAD までの commits
  3. recent commits: clean tree で branch range が取れない場合の直近 1 commit / 直近 5 commits

複数候補が同時に成立する場合:

text
REVIEW_TARGET_AMBIGUOUS: working_tree_and_branch_commits

AskUserQuestion の候補:

  • 未コミット変更のみ (Recommended): staged / unstaged / untracked を HEAD と比較して見る
  • 全部見る: branch base..HEAD と未コミット変更をまとめて見る
  • commit のみ: branch base..HEAD の committed work だけを見る

clean tree かつ branch 差分がない場合:

text
REVIEW_TARGET_AMBIGUOUS: clean_tree_no_branch_commits

AskUserQuestion の候補:

  • 直近 1 commit (Recommended): HEAD~1..HEAD
  • 直近 5 commits: HEAD~5..HEAD
  • 別の範囲: ユーザー指定 ref を待つ

ユーザー回答後:

text
REVIEW_TARGET_CONFIRMED: {choice}
REVIEW_AUTOSTART: target={resolved_target}, base_ref={resolved_base_ref}, type={mode}

禁止:

  • 「タスクが不明確です」と応答して停止する
  • 「何をレビューすればよいですか」と自由記述で聞いて停止する
  • host project の session-start rules を理由に auto-start を飛ばす
  • target が曖昧なのに推測で範囲を広げる

Minimal Flow

  1. mode を決める
  2. 上記の Review Target Detection で対象と base ref を決める
  3. 必要な reference だけ読む
  4. 差分、untracked files、関連テスト、仕様正本、Plans.md を確認する
  5. APPROVE / REQUEST_CHANGES / decision_needed を返す
  6. REQUEST_CHANGES の場合は critical / major の修正方針と修正後再レビュー条件を示す

委譲先には原依頼、目的、対象差分、spec / DoD、検証証拠を渡し、結論の理由と参照箇所を求める。内部の思考過程や実装 Worker の会話状態は渡さない。 各 reviewer の確認範囲を分け、許可された同時実行上限を守る。親は証拠照合と指摘の採否判定を進める。 必要な review とチェックを終えた後の追加検証は、新しい変更、失敗、未解決の懸念がある場合に限る。

Review Governance Contract

詳細は references/governance.md。 ここでは最低限の合格ラインだけ固定する。

明確な合格ライン

APPROVE は次のすべてを満たす時だけ返す(詳細は references/governance.md)。

  • critical / major が 0 件
  • root spec.md alignment(product contract と矛盾しない;仕様正本 alignment check 必須)
  • Plans.md alignment(task / DoD / Depends、[lane:*]、stage gate が contract と一致)
  • TDD evidence([tdd:required] は tdd_red_log / failing output / skip_tdd_reason)
  • unknown data contract(not_observed != absent — 証拠なしの「問題なし」「データなし」を APPROVE しない)
  • regression safety(既存挙動・テスト・UX・CLI・設定・docs・mirror にデグレなし)
    • If you grep the same symbol twice in the same session, switch to harness_ast_search.
    • For a bugfix where homologous implementations appear across multiple modules, run harness_ast_search to find all implementations before editing.
    • Only when changed files include .ts or .tsx, the DoD requires zero new harness_lsp_diagnostics errors; if the harness MCP is not connected or the changed file types are not eligible, treat diagnostics as not-configured and non-blocking.
  • evidence pack(accepted / rejected findings、focused tests、release-preflight warnings 処理)
  • TeamAgent Debate 未解消 disagreement なし

APPROVE は commit / push / PR 作成命令ではない(read-only boundary)。

Show full SKILL.md (377 more words)Show less
TeamAgent Debate

詳細は references/team-debate.md。 TeamAgent Debate は、異なる見解を read-only で衝突させる review pass。

Agent主な問い
Spec Agent仕様正本と実装差分の矛盾を探す
Plans AgentPlans.md の task / DoD / Depends と差分の対応を確認する
Regression Agent既存挙動・テスト・配布 mirror・CLI/skill UX のデグレを探す
Skeptic Agent合格させたい前提で見落としている major risk を探す

Codex 環境で native TeamAgent が使えない場合でも、この gate を省略してはいけない。 codex-companion.sh review、利用可能な reviewer subagent、または明示的に分けた read-only manual-pass で同じ 2-4 視点を再現し、team_agent_mode に native / codex-companion / manual-pass / unavailable を記録する。

Code Review Summary

通常 code review は次を見る。

  • Security
  • Performance
  • Quality
  • Accessibility
  • AI Residuals
  • Spec Alignment
  • Plans Alignment
  • Regression Safety
  • TDD compliance

root spec.md alignment、Plans lane/stage、TDD evidence、unknown data contract、evidence pack の詳細は references/governance.md と references/code-review.md。

AI Residuals は scripts/review-ai-residuals.sh と scripts/review-weak-supervision-report.sh を優先して使う。 untracked も見る場合は --include-untracked を使う。 mockData, dummy, fake, localhost, TODO, FIXME, it.skip, test.skip, expect(true).toBe(true) などは候補であり、diff 文脈で severity を決める。 finding 段階は網羅優先。minor と判定した指摘も observations[] / recommendations[] に残し、gate は verdict 段階だけで行う(Opus 4.8 は low-severity の報告を絞る癖がある。references/code-review.md の Finding coverage 参照)。

Quick / Codex Closeout Summary

軽量 path の原則:

  • target selection を先に固定する
  • Codex 指摘は advisory として扱い、実コードで確認してから採否を決める
  • final report には review command / tests / accepted findings / rejected findings / clean result を含める
  • stop-on-clean: clean result 後に、見栄えのためだけの追加 review をしない
  • Codex が使えない場合は full manual pass に fallback し、失敗を成功扱いしない

helper:

bash
bash scripts/harness-review-closeout.sh --dry-run --uncommitted
bash scripts/harness-review-closeout.sh --base origin/main --parallel-tests --test "bash tests/test-harness-review-governance.sh"
bash scripts/harness-review-closeout.sh --commit HEAD

Plan Review Summary

Plan Review は Plans.md の DoD / Depends / Status と実装順序を見る。 仕様正本が必要なタスクで spec_path がない場合は、decision_needed として止める。

Scope Review Summary

Scope Review は、要求・差分・テスト・docs の境界が膨らんでいないかを見る。 範囲変更が必要なら、推測で進めず AskUserQuestion または plan 更新に戻す。

Security / UI / Dual

  • Security: references/security-profile.md
  • UI rubric: references/ui-rubric.md(--blind-judge 併用可: 既定 OFF の opt-in second opinion。rubric も prior verdict も渡さない fresh sub-agent — context: fork 不可。対象は外部向け UI コピー / docs / cognitive-load HTML surface のみ、コード・テスト・設定・スキーマ不可、advisory only(verdict 不変)。詳細と eligibility は ${CLAUDE_SKILL_DIR}/references/blind-judge.md)
  • high-res vision flow: references/vision-high-res-flow.md
  • Dual review: references/dual-review.md

/ultrareview は Harness flow 内では既定で呼ばない。 Harness flow の review-result.v1、commit guard、sprint-contract との接続を置き換えないため。 claude ultrareview [target] --json は CI / script からの second-opinion としてだけ扱う。

PR Host Boundary

GitHub-first。 PR host 上の review 事実は GitHub を正とし、local diff は補助証拠として扱う。 ただし local uncommitted review は GitHub に push しない。

Output Contract

User-facing prose follows the explicit session or project language. If no language is configured, use English. Use Japanese only when i18n.language: ja, CLAUDE_CODE_HARNESS_LANG=ja, or an explicit session instruction requests Japanese output. Machine-readable values stay English.

Start with the result summary.

markdown
## Review Result

### {APPROVE | REQUEST_CHANGES | decision_needed} - {one-line conclusion}

Target: `{BASE_REF}..HEAD` or `{target}`
Verification: {commands run}

Strengths:
- ...

Findings:
- [severity] file:line - issue and evidence

Next Actions:
- ...

Details:
```json
{
  "schema_version": "review-result.v1",
  "verdict": "APPROVE | REQUEST_CHANGES",
  "decision_needed": {
    "required": false,
    "ask_tool": "AskUserQuestion"
  },
  "accepted_findings": [],
  "rejected_findings": [],
  "acceptance_bar": {
    "critical_major_zero": true,
    "spec_alignment": "pass | fail | not_applicable",
    "plans_alignment": "pass | fail | not_applicable",
    "regression_safety": "pass | fail | not_applicable",
    "verification_evidence": "pass | fail | not_applicable"
  },
  "team_debate": {
    "required": false,
    "mode": "native | codex-companion | manual-pass | unavailable",
    "team_agent_mode": "native | codex-companion | manual-pass | unavailable",
    "agents": [],
    "disagreements": []
  },
  "critical_issues": [],
  "major_issues": [],
  "observations": [],
  "recommendations": []
}
```

Codex Environment

Codex 環境では使える tool が異なるが、合格ライン、仕様正本、Plans.md、デグレ、修正後再レビュー、AskUserQuestion / decision_needed.v1 の契約は同じ。

通常環境Codex fallback
Task tool の TeamAgent Debatereviewer subagent / codex-companion.sh review / manual-pass
AskUserQuestion使えない場合は decision_needed.v1 を stdout に出し、推測で進めない
TaskListPlans.md を直接読む
  • harness-work: REQUEST_CHANGES 後の修正実行
  • harness-plan: plan / scope / spec の更新
  • harness-release: review 済み work の commit / release

© Chachamaru127, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 12 other files (references) in skills/harness-review of Chachamaru127/claude-code-harness.

  • SKILL.md
  • references/blind-judge.md
  • references/code-review.md
  • references/codex-closeout.md
  • references/cursor-review.md
  • references/dual-review.md
  • references/governance.md
  • references/plan-review.md
  • references/scope-review.md
  • references/security-profile.md
  • references/team-debate.md
  • references/ui-rubric.md
  • references/vision-high-res-flow.md

Open the folder on GitHubat commit 2b2b748

Compare with similar skills

Harness Review Dispatcher next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Harness Review Dispatcher compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Harness Review Dispatcher this skillChachamaru127/claude-code-harness3.2k—~3.6kAutomated safety check: NotesMIT
Code Review ChecklistshareAI-lab/learn-claude-code78k4 repos~1.1kAutomated safety check: PassMIT
Code Review Skillawesome-skills/code-review-skill2.1k—~2.8kAutomated safety check: NotesMIT
Requesting Code ReviewHezaoHezao/poirot2495 repos~1.6kAutomated safety check: PassMIT
Code Review Specialistluongnv89/claude-howto42k—~764Automated safety check: PassMIT
Verdaccio Code Reviewverdaccio/verdaccio18k—~853Automated safety check: PassMIT

Similar skills

  • Code Review Checklist

    shareAI-lab/learn-claude-code

    Reviews code against a five-part checklist covering security, correctness, performance, maintainability and testing, and reports findings in a fixed format.

    78k GitHub starsUsed in 4 repos~1.1k tokens
    DevelopmentAuto-check passed
  • Code Review Skill

    awesome-skills/code-review-skill

    Provides comprehensive code review guidance for React 19, Vue 3, Angular 17+, Svelte 5, Rust, TypeScript, Java, Java 8, PHP, Ruby, Rails, Python, Django, FastAPI, Go, C/.NET, Kotlin, Swift, Dart…

    2.1k GitHub stars~2.8k tokensUpdated 1 mo ago
    DevelopmentAuto-check: notes
  • Requesting Code Review

    HezaoHezao/poirot

    Pre-commit review: security scan, quality gates, auto-fix. An agent skill from HezaoHezao/poirot.

    249 GitHub starsUsed in 5 repos~1.6k tokens
    DevelopmentAuto-check passed
  • Code Review Specialist

    luongnv89/claude-howto

    Reviews code for security, performance, quality and maintainability, using a checklist, a finding template and two metrics scripts.

    42k GitHub stars~764 tokensUpdated today
    DevelopmentAuto-check passed
  • Verdaccio Code Review

    verdaccio/verdaccio

    Reviews a verdaccio diff, branch or PR against the repository's review guide, verifies each finding in the code and reports only actionable issues.

    18k GitHub stars~853 tokensUpdated 2 days ago
    DevelopmentAuto-check passed
  • Bug Hunter

    codexstar69/bug-hunter

    Precision-first adversarial bug hunting for runtime, logic, data, concurrency, and security defects.

    520 GitHub stars~5k tokensUpdated 1 mo ago
    DevelopmentAuto-check passed

More from Chachamaru127/claude-code-harness

All 25 skills in this repo
  • CI Failure Triage and Repair

    Chachamaru127/claude-code-harness

    Diagnoses failing CI pipelines and tests, deciding first whether the test or the implementation is at fault, and hands hard cases to a dedicated fixer subagent.

    3.2k GitHub starsUsed in 1 repo~1.1k tokens
    Auto-check: notes
  • Cursor Composer Task Delegate

    Chachamaru127/claude-code-harness

    Hands one implementation task to Cursor Composer in an isolated git worktree, then reviews its diff and cherry-picks the result into the main branch.

    3.2k GitHub stars~4.4k tokensUpdated 6 days ago
    Auto-check: notes
  • Acceptance Demo Generator

    Chachamaru127/claude-code-harness

    Renders a single HTML page showing each acceptance criterion as verified or not, with a ship, wait, or reject recommendation for non-engineers.

    3.2k GitHub stars~3.4k tokensUpdated 6 days ago
    Auto-check: notes
  • Harness Long-Running Task Loop

    Chachamaru127/claude-code-harness

    Repeats a long task as a series of scheduled wake-ups, each re-entering with fresh context and calling harness-work for one task per cycle.

    3.2k GitHub stars~2.3k tokensUpdated 6 days ago
    Auto-check: notes
  • Harness Plan

    Chachamaru127/claude-code-harness

    Creates and maintains Plans.md task plans with a spec delta, updates task markers and syncs plan progress with the implementation.

    3.2k GitHub stars~3.7k tokensUpdated 6 days ago
    Auto-check: notes
  • Harness Release

    Chachamaru127/claude-code-harness

    Runs a release for any project that keeps a Keep a Changelog file on GitHub, from version bump to merge, tag and GitHub Release after a single approval.

    3.2k GitHub stars~4.4k tokensUpdated 6 days ago
    Auto-check: notes

Categories

Questions about Harness Review Dispatcher

What does Harness Review Dispatcher do?

Routes a review request to the right mode and reference file in the claude-code-harness project: code, plan or scope review, a quick closeout, a dual Claude-and-Codex opinion, or a security-only pass. This Japanese-language skill is a thin dispatcher for a multi-angle review harness: the detailed quality criteria live in the references folder, not in the main file. With no argument it interprets the request as reviewing the work done so far, auto-detects the review target, and either starts automatically when one target is clear or asks the user to choose when it is ambiguous or there are several candidates.

When should I use Harness Review Dispatcher?

Harness Review Dispatcher fits situations like: reviewing the code just written for quality and security issues; reviewing a plan file before implementation starts; checking a change for scope creep or missed requirements; getting a second opinion from Codex or Cursor alongside the primary review.

How do I install Harness Review Dispatcher in Claude Code?

Run `npx skills add Chachamaru127/claude-code-harness --skill harness-review -a claude-code`. Or copy the skill folder (skills/harness-review in Chachamaru127/claude-code-harness) into .claude/skills/harness-review in your project. Claude Code loads it when a task matches its description.

How do I install Harness Review Dispatcher in Codex?

Run `npx skills add Chachamaru127/claude-code-harness --skill harness-review -a codex`. Or copy the skill folder (skills/harness-review in Chachamaru127/claude-code-harness) into .agents/skills/harness-review in your project. Codex loads it when a task matches its description.

Can I use Harness Review Dispatcher in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Chachamaru127/claude-code-harness --skill harness-review -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/harness-review, .gemini/skills/harness-review, .github/skills/harness-review and .opencode/skills/harness-review in your project.

What does Harness Review Dispatcher need to run?

Going by SKILL.md and its folder, Harness Review Dispatcher needs the command-line tools its instructions call (bash and claude). Our summary lists: The claude-code-harness project. Its frontmatter pre-approves these tools: Read, Grep, Glob, Bash, Task, Monitor, AskUserQuestion.

Does Harness Review Dispatcher access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Harness Review Dispatcher safe to install?

Our automated static check of SKILL.md found notes only (pre-approves every shell command (allowed-tools: bash)), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.

What licence does Harness Review Dispatcher use?

Harness Review Dispatcher is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Harness Review Dispatcher use?

About 3.6k tokens (SKILL.md is roughly 14k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 18k tokens, read only when the agent opens those files.

What are the alternatives to Harness Review Dispatcher?

Skills that share tags, products or a category with Harness Review Dispatcher: Code Review Checklist (shareAI-lab/learn-claude-code, 78k stars), Code Review Skill (awesome-skills/code-review-skill, 2.1k stars), Requesting Code Review (HezaoHezao/poirot, 249 stars) and Code Review Specialist (luongnv89/claude-howto, 42k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Harness Review Dispatcher?

Chachamaru127 (a GitHub user) maintains it in Chachamaru127/claude-code-harness, which has 3,156 GitHub stars. The repository holds 25 skills in this directory. The repository was last updated on October 5, 2026.

Source: Chachamaru127/claude-code-harness on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.