chore: mark 0017 and 0018 HITL gates complete via HOTL subagent tests

Behavioral tests for write-eval, write-skill, and write-docs run via
fresh-context subagents (HOTL). All process steps verified correct.
Caveman defects surfaced during testing logged in 0028 for upgrade-skill.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docs/issues/0028-chunk-3-closure.md

@@ -16,6 +16,8 @@ Close out Chunk 3 once all 42 skills are complete: update the skills index to re
 
 **Behavioral test scope:** All 42 skills (including `write-eval`, `write-skill`, and the 4 preserved skills). The `caveman` skill is exempt — it has no content-generating behavior to verify.
 
+**Known caveman defects (surfaced during 0017 HOTL test, 2026-05-26):** caveman is a pre-standard legacy skill pending adoption via `upgrade-skill`. Two defects to fix at that time: (1) missing `metadata.category: cross-cutting` in frontmatter — write-eval cannot compute output path without it; (2) `"be brief"` trigger is over-broad — fires on one-shot brevity requests, not just persistent mode activation. Negative test cases documenting the correct boundary are captured in the HOTL test output.
+
 **LESSONS.md:** Extract any cross-session learnings from Chunk 3 implementation and add entries per the LESSONS.md format. Three or more observations on the same pattern graduate to the relevant standing file.
 6. Review `docs/notes/skill-implementation-workflow.md` — verify the conventions are still accurate; update any entries that changed during implementation.