본문으로 건너뛰기
cccskills
무료GitHub에 공개

review-local

Review local git changes from 8 expert perspectives by spawning the matching reviewer subagents in parallel. Produces a consolidated report with Critical/Important/Nice-to-have severity. Lightweight pre-commit or pre-push quality gate — no PR or branch push required. Use when the user asks to review local changes, check changes before committing, get a team review of working tree changes, or run a pre-commit review. Trigger phrases include "review local", "review my changes", "review local changes", "pre-commit review", "review before commit", "review before push", "team review my changes", "check my changes", "review working tree", "local code review", "review diff", "review my diff".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hideki5123/agent-skill-set

실행 점수 미검증

설치 방법 보기

포함된 파일(19)

  • SKILL.md7.6 KB
  • agents/architecture.md2.5 KB
  • agents/bdd.md2.7 KB
  • agents/devils-advocate.md3.4 KB
  • agents/idempotence.md2.9 KB
  • agents/infrastructure.md2.9 KB
  • agents/qa.md2.7 KB
  • agents/security.md2.7 KB
  • agents/tdd.md2.6 KB
  • references/perspectives/bdd.md1.5 KB
  • references/perspectives/devils-advocate.md1.4 KB
  • references/perspectives/idempotence.md1.5 KB
  • references/perspectives/infrastructure-design.md1.5 KB
  • references/perspectives/qa-qc.md1.4 KB
  • references/perspectives/security.md1.4 KB
  • references/perspectives/software-architecture.md1.4 KB
  • references/perspectives/tdd.md1.5 KB
  • references/report-template.md1.4 KB
  • references/synthesis-protocol.md1.4 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Local Review (orchestrator)

Reviewer judgment lives in eight subagents under review-local:perspectives:*. This skill orchestrates them in parallel, synthesizes findings into a severity-classified report, and (optionally) writes the report to a timestamped file.

Reviewer subagents (spawned in parallel)

SubagentLens
review-local:perspectives:architectureSoftware architecture, SOLID, modularity
review-local:perspectives:securityInput validation, injection, authz, secrets, crypto
review-local:perspectives:infrastructureDeployment, observability, resilience, scale
review-local:perspectives:qaTestability, coverage, edge cases, regression risk
review-local:perspectives:tddRed/green discipline, scope bound by tests
review-local:perspectives:bddUser-observable behaviour, ubiquitous language
review-local:perspectives:idempotenceRetry safety, dedup, delivery semantics
review-local:perspectives:devils-advocateThe case against merging

Each subagent is a focused reviewer that reads the diff + file context and returns a structured review block. The reviewers do not spawn further subagents — Claude Code does not allow nested subagent spawn.

Arguments

ArgumentDefaultDescription
--staged-onlyfalseOnly review staged changes (git diff --cached)
--filenoneWrite report to review-local-YYYY-MM-DD-HHmm.md
--lensallComma-separated subset of perspective names (e.g., architecture,security,qa). Default: spawn all 8
--max-files30Soft limit on changed-file count. Over this, the skill samples (highest-churn files first) rather than asking the user. Set to 0 to disable the cap

Workflow

1. Collect Changes ──► 2. Read Context ──► 3. Parallel reviewer subagents ──► 4. Synthesize ──► 5. Report

Phase 1: Collect changes

git diff --cached                   # staged (always)
git diff                            # unstaged (skip if --staged-only)
git status --porcelain | grep '^??' # untracked (skip if --staged-only)

If all are empty: "No local changes detected. Nothing to review." → exit.

For untracked files, read full file content as a pseudo-diff.

Phase 2: Read context

For each changed file, Read the full file so each reviewer has surrounding context (not just hunk lines). Skip deleted files and binary files.

Large changesets (>--max-files): sample by churn — keep the top --max-files files by git diff --stat | sort -k3 -nr. Note the sampled vs total file count in the final report. Do not ask the user whether to proceed — the soft cap is configured by --max-files, and the default 30 already covers most local-changes flows.

Phase 3: Parallel reviewer subagents

For each lens in --lens (default: all eight), spawn the matching subagent via the Agent tool with subagent_type set to the scoped identifier — review-local:<lens>:

Lenssubagent_type
architecturereview-local:architecture
bddreview-local:bdd
devils-advocatereview-local:devils-advocate
idempotencereview-local:idempotence
infrastructurereview-local:infrastructure
qareview-local:qa
securityreview-local:security
tddreview-local:tdd

Pass each:

  • The combined diff from Phase 1.
  • The file context from Phase 2 (path → content map).
  • Note of any sampled-out files.

Run them in parallel (multiple Agent tool calls in a single message).

Each reviewer returns a structured review block matching its lens's output format. The reviewers are read-only — they do not edit files.

Phase 4: Synthesize

Read references/synthesis-protocol.md for the full protocol. Briefly:

  1. Collect each subagent's block.
  2. Categorize each finding by severity: Critical / Important / Nice-to-have.
  3. Deduplicate overlapping findings — when two lenses point at the same file:line, merge into one row that lists both lenses and keeps the higher severity.
  4. Order Critical → Important → Nice-to-have within each section.

Phase 5: Present the report

Read references/report-template.md for the exact format. Print the report inline. If --file is set, also write review-local-YYYY-MM-DD-HHmm.md in the current directory and confirm the path.

Behavior Scenarios

Scenario: Full eight-lens review of local changes
  Given the user has local changes (staged, unstaged, or untracked)
  When /review-local is invoked
  Then the skill spawns 8 reviewer subagents in parallel
       (review-local:perspectives:architecture, security, infrastructure,
       qa, tdd, bdd, idempotence, devils-advocate)
  And synthesizes their structured findings into a single severity report

Scenario: Staged-only review
  Given the user has staged changes ready to commit
  When /review-local --staged-only is invoked
  Then only the staged diff is sent to the reviewer subagents

Scenario: Subset of lenses
  Given the user only wants security + qa coverage
  When /review-local --lens=security,qa is invoked
  Then only those two subagents are spawned, and the report only contains
       findings from those lenses

Scenario: Write report to file
  Given the user wants a persistent record
  When /review-local --file is invoked
  Then the report is written to review-local-YYYY-MM-DD-HHmm.md
       and the path is confirmed

Scenario: No local changes
  Given the working tree is clean
  When /review-local is invoked
  Then the user is told "No local changes detected" and the skill exits

Scenario: Large changeset
  Given more than 30 files are changed
  When /review-local is invoked
  Then the skill samples the top 30 files by churn, notes the sampling
       in the report header, and does not pause for user confirmation

Why this is an orchestrator (not a single reviewer subagent)

Each of the eight lenses has a distinct system prompt and output format. The cleanest mapping is one subagent per lens. Spawning them must happen from a context that can spawn subagents — i.e., the main session, via this skill — because subagents cannot spawn further subagents in Claude Code. So:

  • The judgment per lens → eight subagents.
  • The parallel-spawn + synthesis + report → this skill.

References

  • references/perspectives/ — Individual perspective definitions. These also serve as the source for each reviewer subagent's system prompt (the agent files in agents/ carry the same content packaged as subagent definitions). Agent files must sit directly in agents/ — Claude Code does not discover them in a subdirectory, and a plugin whose agents are nested registers zero of them.
  • references/synthesis-protocol.md — How to merge, deduplicate, and categorize findings across lenses.
  • references/report-template.md — Exact markdown format for the consolidated report.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

adversarial-panel

무료

Run a multi-model adversarial mutual review with facilitator synthesis: 2+ panelists (different model families where possible — use codex-server / codex-cli for the external-family seat) independently answer, adversarially critique each other across rounds, and the main session acts as facilitator, synthesizing agreements, live disagreements, and calibrated confidence into one answer. Use ONLY on high-stakes, hard-to-reverse claims, designs or decisions, and only when the user EXPLICITLY asks to debate or adversarially review with multiple models — "discuss with Opus and Codex", "red-team this", "devil's advocate", 敵対的レビュー, 相互レビュー, 多モデル討論, 議論させて. A full run costs ~6 subagent/CLI dispatches: state the panel size and round count and get a go-ahead before launching. Where the Workflow tool is available, run the panel as a Workflow (judge-panel / adversarial-verify) instead of ad-hoc fan-out. Do NOT self-trigger on a bare "are you sure?" — that is a cue to OFFER the panel, not to run it.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hideki5123/agent-skill-set22026년 10월 21일 업데이트

agent-swarm

무료

Orchestrate multiple coding-agent sessions (Claude Code, Codex, Gemini) over the mcp_agent_mail MCP server, observe their progress on a recurring auto-loop, and prevent inconsistencies (file/edit conflicts and contradictory decisions) between concurrent tasks. Auto-manages the agent-mail server, derives the shared room key from the git repo, and runs either as the orchestrator/observer or as a participating worker. Handles the operational details (server startup, room key, reservation conventions) so the user does not have to remember them. Use when coordinating several agent sessions on different tasks at once, watching a swarm of sessions for conflicts, or wiring already-running sessions into a shared coordination room. Trigger phrases: "agent-swarm", "/agent-swarm", "orchestrate agents", "orchestrate multiple claude sessions", "coordinate sessions", "watch for conflicts between agents", "observe the swarm", "prevent inconsistency between sessions", "multi-agent observe", "複数セッションを協調", "エージェント協調", "セッション間の不整合監視", "スウォーム監視", "複数のClaudeを束ねて".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hideki5123/agent-skill-set22026년 10월 21일 업데이트

agents-init

무료

Generate AGENTS.md (cross-tool agent context — Codex CLI, Aider, Cursor, etc.) and wire CLAUDE.md as a symlink so Claude Code reads the same file. Delegates content generation to Claude's built-in /init when needed, then renames + symlinks. Idempotent. Triggers — /agents-init, "generate agents.md", "create agents.md", "share claude and codex context", "set up agents file", "init agents", "agents-init", "cross-tool agent context", "wire claude and agents".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hideki5123/agent-skill-set22026년 10월 21일 업데이트

chrome-use

무료

Drive an already-running, LOGGED-IN Chromium browser on macOS via AppleScript — navigate, click, fill forms, scroll, and extract page content from the user's REAL session, with the browser left open and NO profile copy and NO restart. Use when the task needs the existing logged-in session (cookies, auth, current tabs) reused as-is. Trigger phrases: "chrome-use", "ブラウザから取得", "ログイン済みの Chromeで", "ブラウザを操作して", "このページ読んで", "Xのスレッド読んで", "ログインしたまま取得", "自分のセッションでスクレイプ", "read this page in my browser", "scrape with my session", "drive my chrome", "execute JS in my (logged-in) browser", "automate my logged-in browser". Chromium family (Chrome default; Brave/Edge/Arc/Vivaldi via --app). NOT for isolated/throwaway automation where login reuse is unneeded — use playwright-cli for that. macOS only.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hideki5123/agent-skill-set22026년 10월 21일 업데이트

codex-cli

무료

Run OpenAI Codex CLI from the terminal for batch / one-shot `codex exec` invocations with output captured to an `-o` file. Use for fire-and-forget runs, scripted CI, code review batches (`codex review`), and resuming a specific session by id (`codex exec resume`). Always use this skill instead of `mcp__codex__codex` and `mcp__codex__codex-reply` MCP tools — CLI is faster and more capable. For streaming UX, multi-turn dialogues that need live state, structured (JSON schema) output, attaching images, or programmatic event handling → **use codex-server instead**. codex-server uses the user's ChatGPT subscription via Codex App Server and decoupled async invocation that bypasses the Bash 2-minute timeout entirely. Trigger patterns (batch-only — chat/streaming triggers belong to codex-server): codex-cli / codex CLI / codex exec / codex review / codex resume / codex fork / batch + {codex, gpt} / one-shot + {codex, gpt} / non-interactive + codex / codex {to a file, -o, output file, captured output} / CI + {codex, gpt} / scripted + {codex, gpt} / fork codex session / resume codex session by id

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hideki5123/agent-skill-set22026년 10월 21일 업데이트

codex-server

무료

Run OpenAI Codex via the Codex App Server (JSON-RPC-over-stdio) for streaming multi-turn chat sessions with persistent threads, structured output, image input, and rich event streams. Default entry point for ChatGPT/GPT/Codex-like conversations from the terminal. Uses the user's ChatGPT subscription (Plus / Pro / Team) exclusively via `codex login` — never consumes OPENAI_API_KEY billing, by design. Implementation: TypeScript on deno with npm:@openai/codex-sdk; spawns the existing system codex binary. Decoupled async invocation: every turn returns turn-id in <1 s; the actual SDK work runs in a detached worker that streams to per-turn files under ~/.codex-server/turns/<turn-id>/ — Bash's 2-min timeout never applies. Prefer this over codex-cli for streaming UX, multi-turn dialogues that need live state, structured (JSON schema) output, attaching images, or programmatic event handling. Use codex-cli only for one-shot batch invocations that pre-capture to an -o file. For OpenAI Platform API calls specifically (embeddings, image generation, audio, moderation, batches, files) — not open-ended chat — use openai-cli instead; that path bills OPENAI_API_KEY, this one uses the ChatGPT subscription. Trigger patterns (match any variation): codex / codex-server / codex app server / codex app-server / openai / open ai / OpenAI / chatgpt / chat gpt / ChatGPT / GPT / ask gpt / chat with gpt / gpt session / codex session / streaming chat / multi-turn chat / multi turn chat / second opinion / cross-check / cross check / ai pair programming / another AI / external AI / ask another model / get GPT's take / what does openai think / show me what gpt thinks / structured output from gpt / json schema gpt / resume codex thread / continue codex conversation / continue codex session / /codex-server

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

hideki5123/agent-skill-set22026년 10월 21일 업데이트

hideki5123의 스킬 모두 보기

이 스킬의 문제 신고