Recommends harness improvements
Ranks skill, sub-agent, and rule recommendations per repo by stack match, security posture, and observed effectiveness on similar repos.
Recommends stronger skills, sub-agents, and rules for each repo, turns outperforming prompt patterns into skills the rest of the org can adopt, and coaches each engineer on their own sessions.
Ranks skill, sub-agent, and rule recommendations per repo by stack match, security posture, and observed effectiveness on similar repos.
Surfaces where skill or plugin combinations conflict with each other and quietly degrade output — the kind of drift that doesn't show up in a PR review.
Codifies outperforming prompt patterns into installable custom skills, and surfaces per-repo CLAUDE.md/AGENTS.md/Cursor-rules diffs grounded in actual session outcomes.
A weekly per-engineer retro on what worked and what context was missing, grounded in that engineer's own sessions. The engineer sees it before their manager does.