# Process-Input Quality Checklist

Score the project resource records and `docs/inputs/<project-slug>/index.json` against these 9 items before handoff.
**Pass criterion: 9/9.** Any failed item MUST be fixed and re-scored.

1. **Project slug honored.** All reads/writes use `docs/inputs/<project-slug>/...`; no global fallback exists.
2. **Index valid.** `index.json` is valid pretty JSON with `project` and `inputs`; processed entries include `src`, `md`, and `status`.
3. **Every input accounted for.** Each manifest item has a resource entry with `ready`, `failed`, or `ignored`; none are silently missing.
4. **Source-based reuse.** Ready resources with matching `src` were reused without re-transcription or record rewrite.
5. **Resource records correct.** New ready resources wrote one `.md` each under `docs/inputs/<project-slug>/transcriptions/`.
6. **Frontmatter complete.** Each resource record has the fields required by `templates/input-record.md.tmpl`.
7. **Zero fabricated/paraphrased content.** Audio transcripts are MCP output verbatim; text/inline content remains untouched.
8. **Discovery input boundary held.** No audio/text files were created under `docs/discovery/<project-slug>/inputs/`.
9. **No auto-chaining.** No downstream skill (`requirements`, `hla`, ...) was started; `/machine-discovery:requirements <project-slug>` was only suggested.
