What problem does it solve? After code is written, residual bugs, undecided choices, and unverified paths often remain hidden. This Skill runs a structured after-the-fact review that finds what an implementation left behind and ends with concrete, minimal proposed fixes instead of vague impressions. ## Core Features & Use Cases - Three-pass review: implementer self-analysis grounded in file:line evidence, an independent solutions agent that verifies findings and proposes minimal fixes, and a critical-feedback agent that stress-tests each fix for over-reach, under-reach, and regressions. - Classified findings: every suspected issue is labeled CONFIRMED, PARTIAL, or NOT-A-BUG with a reproducible trigger, and each fix gets an apply / apply-with-changes / reject / document-only verdict. - Durable report: results are written to .plans/<initiative>/post-implementation-review.md alongside the brainstorm and plan, so declined fixes stay recorded and nothing is silently lost. - Use Case: You just finished a shell-script change touching process-control code and only verified one path. Run this review to have three independent passes confirm or dismiss your suspicions and deliver exact replacement code for each real defect. ## Quick Start Review what we just built and propose fixes for anything the implementation left behind.
