spec-distillerlisted

Turn already-extracted YouTube artifacts (the JSON produced by the youtube-artifact-collector skill) into a structured Module→Feature→Requirement document, with each requirement traced back to a transcript segment. Use when the user wants requirement extraction, feature/product discovery, business-rule extraction, or a spec/requirements document FROM existing artifacts — e.g. "extract requirements from this artifact", "what features does this video demonstrate", "build a requirements doc from this collection". Runs offline by default (Claude-native, no API key); an optional OpenAI engine is available via `--engine openai`. This consumes Skill 1 JSON, not URLs — it does not download videos or transcripts.
alikmlsipahi/youtube-to-spec · ★ 0 · Data & Documents · score 62

Install: claude install-skill alikmlsipahi/youtube-to-spec

# spec-distiller Skill 2 of the YouTube intelligence pipeline — a **consumption-layer** analysis. It reads the lossless artifacts produced by `youtube-artifact-collector` (Skill 1) and produces a **Module → Feature → Requirement** document. It never touches YouTube; the only input is artifact JSON, and the only coupling to Skill 1 is the artifact's `schema_version` (read defensively). ## When to use this skill Use it when the request is to **analyze already-collected artifacts**: - "Extract the requirements from this artifact." - "What features and business rules does this video/collection demonstrate?" - "Build a requirements/spec document from this collection." If the user instead wants to *collect* metadata/transcripts from URLs or a playlist, that is Skill 1 (`youtube-artifact-collector`), not this skill. ## Engines This skill ships two interchangeable engines that **emit the same output shape**: - **`claude` (default, offline)** — no script runs. You (Claude) read the artifact and the external prompt/template files in-chat and produce the document. No API key required. - **`openai`** — runs `scripts/extract_requirements.py`, which calls the OpenAI API. Selected with `--engine openai`; needs `OPENAI_API_KEY` (see `.env.example`). The prompts and template are **external and swappable** — editing them changes the output for *both* engines without any code change. ## Inputs Either: - a single artifact file — an artifact `.json` produced by Skill 1 (named