Originally created by: Akarsh-Hegde
Supersedes [#103]. This is the same feat/fleet-feedback-loop work with the KAN-141 (deepeval-on-golden-dataset) commits removed — those land separately via the KAN-141-run-deepeval-on-the-golden-dataset branch.
Rebased cleanly onto main (no conflicts). Contains 10 commits:
meridian doctor (9 commits): multi-daemon doctor with rich by-daemon rendering, L1 screenpipe capture preflight checks, per-app a11y checks, config-contract group, worklog/coding-agent/observability groups, root-cause diagnosis + escalation, and --fix tiered auto/guided/manual repair.The following KAN-141 commits were intentionally dropped and are not part of this branch:
feat(classifier): SKILL_PATH override + skill_v1.md A/B candidatefeat(evals): overnight autonomous loop — 7 skill iterations + ETC stage-2 overridefeat(evals): add scripts and configurations for deepeval goldens and sampling sweepsfeat(evals): add new evaluation run and observations for real hand-labeled goldenscargo clippy -- -D warnings && cargo testcd ui && npm run build🤖 Generated with Claude Code
Ticket changed by: Akarsh-Hegde