What problem does it solve? Code reviews often miss real defects because reviewers trust subagent findings without verification, check only the diff instead of all call sites, focus on first-contact paths while ignoring continuation paths, and treat CI gate signals as evidence without confirming the gate exercises the changed code. ## Core Features & Use Cases - Subagent finding verification: Treats code-reviewer subagent output as leads, requiring grep and source-level confirmation before acting on any claim. - Call-site coverage checks: Requires grepping all call sites of new helpers and the old direct calls they replace, not just the branches visible in the diff. - Two-stage protocol review: Provides a first-contact vs continuation checklist for session, conversation, and wire-protocol changes. - CI gate sanity checks: Tests whether a green or red gate actually exercises the changed code, covering Docker builds, lockfile pinning, cached fixtures, and privileged test setup. - Use Case: When reviewing a federation change that introduces an MCP shim, use this Skill to grep every branch of mail and chat tools, verify both first-contact and continuation paths route through the shim, and confirm the Docker e2e gate tests local source rather than a pinned PyPI release. ## Quick Start Load the code-review-discipline skill at the start of my code review assignment and apply its verification checklist to the current diff.
