Full PR lifecycle agent — creates, self-reviews, simplifies, addresses review comments, and loops until the PR is ready for merge. TRIGGER when: user wants to create a PR, review a PR, address PR comments, simplify a PR, make a PR mergeable, or prepare a branch for merge. DO NOT TRIGGER when: user wants to run evals…
Post-eval failure analysis subagent. Partitions failures by optimization level (prompt vs structural) using rule-based heuristics and LLM-based deep analysis. Internal subagent — invoked by the optimization orchestrator after each eval run.
Independent review of a proposed prompt or chain variant against tenant context. Catches guardrail violations the optimization agent might miss (example leakage, placeholder drift, format mismatch, state protocol violations, import safety issues). Internal subagent — invoked by the optimization orchestrator, not…
Instructions for cisco-foundation-ai/fully-automated-prompt-optimization, covering repository guidelines, repository purpose, project structure, build, test, and development and codex workflows.
Instructions for cisco-foundation-ai/fully-automated-prompt-optimization, covering repository guidelines, repository purpose, project structure, build, test, and development and troubleshooting.