Changelog
2026-07-26by Fatih
AI Operations Team skills shipped broken, now fixed
27 AI Ops skills had no frontmatter and no real content, so they never registered as usable slash commands. Also fixes a naming collision between the Evals and Eval agents, and 16 skill files with the wrong filename case.
fix
Fixed
- -AI Operations Team skills: 27 skills across Deploy, Evals, Trace, Guard, Budget, Token, Prompt, Embed, and Rank shipped with no YAML frontmatter and no real step-by-step content, so they were never registering as usable slash commands. All 27 now have proper frontmatter and full instructions.
- -Evals agent skill naming: the Evals agent's skills were named
eval-*, colliding with the unrelated Eval agent from the Data Science team. Renamed toevals-*; two had been accidentally duplicated from the Eval agent's content and now correctly cover LLM evaluation instead of A/B testing. - -16 skills named `skill.md` instead of `SKILL.md` across Form, Draft, and Relay, invisible to any case-sensitive filesystem despite working on case-insensitive local checkouts. Renamed to the correct case.
Added
- -
docs/repomap.md: an orientation map for the repo (100 agents, 421 skills) plus a known-drift log. - -
docs/skill-index.jsonandscripts/gen-skill-index.py: a generated machine-readable skill index, companion to the existingdocs/agent-index.json.