본문으로 건너뛰기
cccskills
무료GitHub에 공개

skill-sync

Sync a skill's content against the upstream library it covers — check whether the API descriptions, version references, and SOURCE: citations inside SKILL.md and its reference files still reflect what the library actually does today. Use when a library has released updates that may have changed the APIs the skill covers, when a skill references an old library version, when asked to sync, refresh, update, or check a skill against upstream docs, when the skill's content may have drifted from current library behavior, or when SOURCE: citations are stale. Accepts a SKILL.md path, skill directory, or plugin directory.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

연동 도구AWS
설치 방법 보기

포함된 파일(5)

  • SKILL.md6.5 KB
  • evals/evals.json2.1 KB
  • references/change-plan-format.md2.3 KB
  • references/report-formats.md2.7 KB
  • references/url-fetch-spec.md4.3 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

<sync_target>$1</sync_target> <invocation_args>$ARGUMENTS</invocation_args>

Argument Contract

InputResolution
Path to a SKILL.md fileOne pipeline for that skill
Path to a skill directory containing SKILL.mdOne pipeline for that skill
Path to a plugin directory containing skills/Glob skills/*/SKILL.md → one pipeline per skill
Anything else or emptySTOP — report ambiguity, ask for a skill path

Pipeline

flowchart TD
    S1["Stage 1 — Token Profile<br>uvx skilllint@latest check SKILL.md<br>wc -w on references/ for size map"] --> S1Q
    S1Q{"Existing file<br>already over SK007?"}
    S1Q -->|"Yes"| S1R["Apply /refactor-skill<br>progressive-disclosure extraction<br>Restart pipeline after"]
    S1Q -->|"No"| S2
    S2["Stage 2 — Parallel Read Agents<br>3 Agent() calls in ONE turn"] --> S3
    S3["Stage 3 — Synthesis<br>Fetch docs index, validate NEW URLs<br>Write change plan to .tmp/scratch/plans/"] --> S3Q
    S3Q{"Change plan has<br>changes?"}
    S3Q -->|"No — all VERIFIED"| Done1(["No changes needed — exit"])
    S3Q -->|"Yes"| S4
    S4["Stage 4 — Clean-Tree Gate<br>git status --porcelain"] --> S4Q
    S4Q{"Output empty?"}
    S4Q -->|"Not empty"| S4Stop(["STOP — dirty working tree<br>List uncommitted files"])
    S4Q -->|"Empty"| S45
    S45["Stage 4.5 — Orchestrator Pre-Fetch<br>Fetch all SOURCE URLs from change plan<br>Write verified content to source-material file<br>URLs that 404 → UNVERIFIABLE, skip"] --> S5
    S5["Stage 5 — Write Agent<br>Agent(skill-content-updater)<br>Receives change plan + source-material path"] --> S5Q
    S5Q{"Post-write skilllint<br>exit code?"}
    S5Q -->|"0"| S6
    S5Q -->|"non-zero"| S5R["Extract over-budget sections<br>to references/<br>Update change plan"]
    S5R --> S5
    S6["Stage 6 — Completion Gate<br>skilllint on ALL modified files"] --> S6Q
    S6Q{"All pass?"}
    S6Q -->|"Yes"| Done2(["Pipeline complete"])
    S6Q -->|"No"| S5

Stage Definitions

Stage 1 — Token Profile

Run uvx skilllint@latest check <skill-path>/SKILL.md (target the file, not the directory — passing a directory silently reports "Total files: 0" and exits 0 without validating anything). Note whether it passes or fails SK007.

To identify the largest reference files before adding content, run wc -w <skill-path>/references/*.md | sort -rn | head -5 and surface the top 3 by word count. This indicates where budget pressure is highest.

Pre-write SK007 branch: if the existing SKILL.md is already over SK007, apply /plugin-creator:refactor-skill (progressive-disclosure extraction) before making content changes. Restart the pipeline after the refactor completes.

Stage 2 — Parallel Read Agents

Dispatch exactly 3 agents in one concurrent fan-out:

  1. plugin-creator:skill-auditor — input: <skill-path>; output: .tmp/scratch/reports/skill-sync-{slug}-completeness-YYYYMMDD.md (read-only)
  2. plugin-creator:skill-content-updater (read role) — upstream drift scan; output: drift report with NEW/STALE/VERIFIED/UNVERIFIABLE verdicts per claim
  3. general-purpose — structure validation; checks progressive disclosure, frontmatter schema, broken reference links; output: structure report

Each agent writes its report to .tmp/scratch/reports/. Report formats: ./references/report-formats.md

Note on drift scanner URL fetching: The skill-content-updater agent fetches the SOURCE: citation URLs already present in the skill, as defined in ./references/url-fetch-spec.md. Do not instruct it to fetch an external docs index — that is the skill-sync-source-validator agent's responsibility in Stage 4.5.

Stage 3 — Synthesis

Orchestrator reads all three reports and writes the change plan to .tmp/scratch/plans/skill-sync-{slug}-YYYYMMDD.md.

If all verdicts are VERIFIED and no structural issues found: write a "no changes needed" change plan and skip Stages 4–6.

Change plan format and synthesis precedence rules: ./references/change-plan-format.md

Stage 4 — Clean-Tree Gate

git status --porcelain

If output is not empty: STOP. List the uncommitted files. Report to the user. Do not proceed until the tree is clean.

Stage 4.5 — Source Validation Agent

Dispatch plugin-creator:skill-sync-source-validator. Pass the change plan path.

The agent validates and pre-fetches all source content before the write agent runs, keeping URL fetching out of the write agent's context and providing a structured fallback chain when primary fetch methods fail. It handles:

  • Validating all NEW URLs against the docs index (llms.txt / sitemap) — downgrades fabricated URLs to UNVERIFIABLE directly in the change plan file
  • Pre-fetching content for every SOURCE URL using the priority fetch chain defined in ./references/url-fetch-spec.md
  • Writing .tmp/scratch/fetched/source-material-YYYYMMDD.md with verified content per change plan entry
  • Updating the change plan header with the source-material file path

Stage 5 — Schema-Aware Write Agent

Dispatch one plugin-creator:skill-content-updater in write role. Pass the change plan path. The change plan header now contains the source-material file path — the write agent reads from that file and does not need to fetch any URLs.

After the agent returns, run uvx skilllint@latest check <modified-skill-path>/SKILL.md. If non-zero: extract the over-budget sections to references/, update the change plan with the extraction directive, and re-dispatch the write agent.

Stage 6 — Completion Gate

Run uvx skilllint@latest check on every file modified in Stage 5. All must exit 0. If any fail: return to Stage 5 with a targeted remediation change plan.

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

Add automated documentation updater to any Claude skill. Creates a Python sync script that downloads upstream docs, processes markdown for AI consumption, and maintains local cache with configurable refresh. Collects template variables, then delegates implementation through 5-phase workflow. Use when adding auto-updating reference documentation to plugins or skills.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Jamie-BitFlight/claude_skills672026년 10월 20일 업데이트

SAM-style feature initiation workflow — discovery through codebase analysis, architecture spec, task decomposition, validation, and context manifest. Use when a user asks to add a feature, plan a feature, or convert an idea into an executable SAM plan.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Jamie-BitFlight/claude_skills672026년 10월 20일 업데이트

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Jamie-BitFlight/claude_skills672026년 10월 20일 업데이트

Browser automation for AI agents using the agent-browser CLI and Playwright. Use when navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, logging into sites, or automating any browser task. Triggers on "open a website", "fill out a form", "click a button", "scrape data", "test this web app", "automate browser actions", or any programmatic web interaction request.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Jamie-BitFlight/claude_skills672026년 10월 20일 업데이트

Runs the description-drift experiment — spawns all Claude Code agents simultaneously to collect self-reported capabilities, then compares them against static frontmatter descriptions to reveal how reliable orchestrator routing based on descriptions actually is. Use when measuring description drift across the agent fleet, re-running the capability collection experiment, analyzing a specific agent's self-reported capabilities, or auditing whether frontmatter descriptions accurately reflect agent behavior.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Jamie-BitFlight/claude_skills672026년 10월 20일 업데이트

Create or adapt Claude Code agent definitions. Use when creating an agent, changing subagent configuration, selecting agent scope, or designing a specialized delegation role.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

Jamie-BitFlight/claude_skills672026년 10월 20일 업데이트

Jamie-BitFlight의 스킬 모두 보기

이 스킬의 문제 신고