Use to close the loop on agent quality — turn a failure signal into a verified fix. Triggers: "improve my agent", "self-improving agent", "auto-tune / iterate on my prompt", "fix my agent from test results" — and production-call bug fixing: "fix this prod call issue", "debug and fix call ID", "reproduce this production bug", "regression test before raising PR". Works on any stack: provider-dashboard agents (VAPI, Retell, ElevenLabs, Bland) AND agents whose config lives in the customer's own repo, database, or prompt registry with the provider agent created at deploy time, or with custom mock servers. Fixed safety invariants (must-fail-first reproduction, attestation, no-prod-in-loop) + a per-project capability manifest declaring where config lives and how to read/render/apply/deploy/verify it.