メニュー
Validate plugin structure, marketplace.json consistency, and changelog format across all plugins
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Produce a source-grounded, cited HTML research report on a question. Gathers a brief (depth, recency, source mix, audience) then launches a bundled Dynamic Workflow that fans out firecrawl retrieval into findings, gates rounds on coverage + contradictions, synthesizes a cited answer, and renders an evolving HTML report. Use for substantive research requests: 'research X', 'zbadaj/zresearchuj X', 'do a deep dive on', 'find out everything about', 'write me a report on', 'compare A vs B with sources', 'what does the evidence say about'. Invoke explicitly as /researcher:research "<question>". Requires Dynamic Workflows enabled + the firecrawl MCP.
Code review focused exclusively on comment quality — not logic, naming, or structure. Use when the user says "/comment-review", "review komentarzy", "sprawdź komentarze", "przejrzyj komentarze", "comment review", or asks to check whether comments in a change are worth keeping. Reviews the current branch diff by default (auto-detects the base branch, or pass --base), or explicit file/dir paths when passed as arguments. Judges every comment against a focused rule set — no code-narration, decisions-only, not too long, no cross-file/doc/spec-id refs (file paths and bare requirement tokens like F1, Q1, R2, §4.1), no banner sections, no change-state/ticket history, no process-narration disguised as a decision, no commented-out code, and nothing that contradicts the code, and no rationale parked on the wrong declaration when it belongs where the behavior lives — and returns a per-comment verdict (KEEP / REMOVE / REWRITE / MOVE) with a concrete suggested fix.
Code review focused on code quality and craft — readability, structure, and style — not correctness, security, or naming. Use when the user says "/quality-review", "review jakości", "sprawdź jakość kodu", "przejrzyj pod kątem jakości", "quality review", "czy ten kod się dobrze czyta", or asks whether a change reads cleanly and is well structured. Especially apt for code freshly generated by a strong model, which is correct but glued together: blocks with no blank lines, helpers ordered randomly, OOP and functional styles mixed, pointless barrel exports, unnecessary casts, and two near-identical functions that should be one. Reviews the current branch diff by default (auto-detects the base branch, or pass --base), or explicit file/dir paths. Returns per-finding severity + a concrete suggested fix, and offers to apply the safe ones.
Orchestrate local-file transcription pipeline — transcribe audio/video files from disk and process them (summarize, extract domain news, or apply custom prompts). Use whenever the user provides paths, globs, or folders pointing at local audio/video files combined with analysis intent. Triggers on: paths ending with audio/video extensions (mp3, m4a, wav, mp4, mov, mkv, etc.) + 'transkrybuj', 'podsumuj nagranie', 'transcribe these files', 'extract action items', 'co jest w tym nagraniu', 'wyciągnij notatki z meetingów', OR folder/glob paths combined with the same intent keywords. Always use this — not local-transcribe or transcript-process separately — when the user wants end-to-end local file analysis. For YouTube URLs instead, use the `yt` skill.
Transcribe local audio/video files to Markdown via ffmpeg + ElevenLabs Scribe. Use whenever the user provides paths to local audio/video files (mp3, mp4, m4a, wav, mov, mkv, etc.) and wants raw transcription without summarization. Triggers on: paths with audio/video extensions + 'transcribe', 'transkrybuj', 'pobierz tekst z nagrania', 'get transcript of this file'. Sub-skill of `local` orchestrator — usually invoked via that. Use directly when the user wants only transcription, no processing.
Process audio/video transcriptions — summarize, extract news/novelties from a specific domain, or apply a custom prompt. Source-agnostic: works with transcripts from yt-transcribe, local-transcribe, or any other producer. Use whenever user has a transcript and wants to do something with it: 'summarize this', 'what's new in AI from this episode', 'extract key points', 'find news about X', 'process this transcript', 'wyciągnij action items', 'podsumuj nagranie'.
| name | verify |
| description | Validate plugin structure, marketplace.json consistency, and changelog format across all plugins |
Run the following checks for every plugin in plugins/*/:
For each plugin directory, verify these files exist:
.claude-plugin/plugin.jsonCHANGELOG.mdREADME.mdReport any missing files.
For each plugin's .claude-plugin/plugin.json, verify:
name, version, description, author, repository, keywordsversion matches semver format (e.g., 1.0.0, 0.1.1)name matches the plugin directory nameRead .claude-plugin/marketplace.json and for each plugin entry:
name matches the plugin's plugin.json nameversion matches the plugin's plugin.json versionsource path points to the correct directory (./plugins/<name>)description is present and non-emptyReport any plugins in plugins/ that are missing from marketplace.json, and any marketplace entries pointing to non-existent plugins.
For each plugin's CHANGELOG.md:
## [Unreleased] section (required by the release script)Print a summary with pass/fail for each check. If all checks pass, confirm everything is consistent. If any fail, list exactly what needs to be fixed.