What problem does it solve? Code review intuition is often generic and misses the specific mistakes that reviewers have repeatedly flagged on a repository. This Skill audits the latest commit (or full branch diff) in the lfx-self-serve repo against a knowledge base of patterns extracted from real PR review comments, so every finding is grounded in an actual documented rule rather than guesswork. ## Core Features & Use Cases - KB-gated findings: Every finding must quote a pattern entry's rule ID and a phrase from its Pattern or Detect clause; unsourced findings are dropped automatically. - Routed pattern loading: Reads only the knowledge-base files relevant to the changed file paths (TypeScript correctness, templates and accessibility, server request handling, observability, Snowflake data access, and more), avoiding wasted context. - False-positive suppression: Applies a known-false-positives list after matching so documented non-issues are removed even when a pattern matches. - Two audit modes: Default mode reviews the latest commit; passing the keyword branch audits the full branch diff against origin/main for the pre-PR sweep. - Use Case: After committing a change to an Angular component and its service in lfx-self-serve, invoke this Skill to receive a markdown report of Critical and Important findings, each citing file, line, confidence score, the KB rule it matched, and the recommended fix. ## Quick Start Ask the AI to run the lfx-self-serve learnings review on the latest commit, optionally adding the keyword branch for a full-branch audit before opening a pull request.
