# Completion audit: AI slop cleanup goal

## Requirement → evidence

- **Create a roadmap for checking all generated course lessons for AI slop.** Evidence: `course/AI-SLOP-AUDIT.md` contains the audit roadmap, scope, checklist, file ledger, findings, fixes, and verification evidence.
- **Use the humanize skill.** Evidence: `course/AI-SLOP-AUDIT.md` records the applied humanize rules: preserve meaning/source links, avoid invented claims, remove chatbot wrappers, inflated claims, fake depth, repeated generic transitions, and noun-heavy phrasing.
- **Audit every relevant course artifact under `course/`.** Evidence: `course/AI-SLOP-AUDIT.md` ledger covers 91 HTML artifacts and 23 Markdown/docs artifacts, with every item marked audited or rewritten.
- **Rewrite detected issues while preserving meaning, links, sources, quizzes, and behavior.** Evidence: 11 high-risk files were rewritten/humanized and listed under `Findings and fixes`; validation still reports 91 valid HTML files and all internal HTML links intact.
- **Do not add unsupported external claims or alter source transcripts.** Evidence: all edited paths are under `course/`; source links point back to existing internal `lessons/`, `types/`, `concepts/`, and `differences/`; no files outside `course/` were edited for the audit.
- **Do not discard user changes or leave TODO placeholders.** Evidence: no source deletion occurred; `course/MATERIAL-TODO.md` has no pending `[ ]` or `[~]` checklist items after the legend; `course/AI-SLOP-AUDIT.md` has all audit checklist items checked.
- **Validate and fix failures rather than stopping at a plan.** Evidence: cleanup was followed by exact-slop scan, roadmap audit, HTML/link validation, Markdown link validation, and HTTP smoke checks; all passed.
- **Verification commands.** Evidence from fresh commands:
  - `bad_exact_hits 0`
  - `todo audit ok: no pending checklist items after legend`
  - `html/link/style ok: 91 html files`
  - `markdown local links ok`
  - HTTP smoke checks returned `200 OK` for `presence/0001-notice-without-attacking.html`, `type-differences/reference-how-to-distinguish-types.html`, and `type-3/reference-type-3-fragmentary.html`.
- **Rendered/smoke inspection of representative lessons.** Evidence: local HTTP server served the representative pages with 200 responses and full HTML bodies after cleanup.
- **No blockers remain.** Evidence: no unmet roadmap items, failed validations, missing links, or missing decisions remain.

Conclusion: every explicit requirement of the active goal is verified by files, audit artifacts, and command output. The goal can be marked complete.