When implementation is cheap but understanding a codebase is not, the bottleneck shifts from writing code to deciding what should change and specifying it clearly. This skill turns that judgment into the deliverable: plans that another agent can carry out without needing the original audit conversation.
What Sets It Apart
- It audits across correctness, security, performance, tests, maintenance, dependencies, developer experience, documentation, and product direction, then rechecks cited evidence to filter out false positives.
- Plans include repository-specific context, verification commands, scope boundaries, and stop conditions. That makes handoffs more reliable than a list of suggestions, especially for less capable executors.
- The workflow continues after planning: an executor can work in an isolated worktree, while review and reconciliation check results and keep the backlog current.
- Source-code changes remain outside the advisor's role. The separation reduces the risk of an audit silently turning into unreviewed implementation, while keeping final merge decisions with the repository owner.
Who It’s For
Great fit for teams that want a strong model to reason about a large or unfamiliar codebase, but prefer cheaper agents or developers to handle scoped implementation. Look elsewhere if you want an agent to directly fix issues, or if your project lacks clear tests and checks that plans can use as verification gates.
