بنقرة واحدة
claude-skills
يحتوي claude-skills على 8 من skills المجمعة من shawnpetros، مع تغطية مهنية على مستوى المستودع وصفحات skill داخل الموقع.
Skills في هذا المستودع
Use when the user asks you to research a topic and wants thesis-grade synthesis, not a bullet-point summary. Activates on "research X", "deep dive on X", "find out about X", "what's actually known about X". Dispatches depth + breadth + counter-argument subagents concurrently, runs a peer-review pass over their findings for hallucination and citation quality, then synthesizes into a structured report with explicit high-confidence / contested / unknown buckets. Output file optionally persisted to a knowledge vault.
Use when tackling a multi-day rewrite, ambitious end-to-end build, or any project that benefits from parallel subagent orchestration with adversarial review. Activates on "boil the ocean", "V3 rewrite", "let's rebuild X", "parallel agents", "multi-wave", or when the user explicitly authorizes large-scope autonomous work. Codifies wave structure, concurrency caps, live-integration discipline, telemetry, and retrospective. NOT for client work or production systems with real blast radius, use a surgical-precision pattern there.
Use when the user wants to document, refresh, or backfill understanding of a project they built - produces a 4-section explanation artifact (what / why / what would break / what learned) by reading the repo first then interviewing only on the gaps. Activated by /explain-project, "help me explain this", "backfill this README", "what did I build here", "turn this into a build-in-public post", or when returning to a stale project after time away.
Promote a substantive insight, synthesis, or analysis generated in chat back into your notes so explorations compound instead of vanishing into transcript history. Lean skill, offers a single file destination and writes. Activated by: /file-insight, 'file this', 'save this analysis', 'promote this to the vault', 'this is worth keeping', or proactively offered when the assistant produces a multi-paragraph synthesis the user acknowledges as valuable.
Universal INGEST operation for a personal knowledge base. Integrates any source material into the vault: media URLs (YouTube/podcasts), existing markdown files, web articles, PDFs, dropped text. Reads source -> extracts key info -> writes summary page -> updates index.md -> touches relevant existing pages -> appends to log.md. Activated by: /ingest, 'ingest this', 'add to vault', 'file this source', 'research this video', 'research this article', or pointing at a file path.
STOP framework: event-driven scenario gate. Enforces BDD scenarios in features.json before implementation (Gate In) and verification after implementation (Gate Out). Optional hooks provide hard enforcement; this skill provides guidance.
Use when ending a work session, the user says done/thanks/that's it/wrapping up, or context is getting long. Also invoked directly via /wrapup. Promotes session history into project pages, updates SESSION-CONTEXT.md + features.json, commits.
Review someone else's pull request. Activated when user asks to review, check, look at, or give feedback on a PR. Fetches the diff, analyzes code quality and correctness, writes inline comments and a summary review in the user's voice. NOT for drafting PRs - this is for reviewing other people's work.