Fix Repository Diagnostics

Run a closed diagnostic loop: discover tooling, collect the baseline, verify findings and select the repair batch, fix root causes, and rerun diagnostics before reporting.

Follow explicit user instructions when they override this skill's defaults. For fix requests, carry the selected batch through verification and reporting without pausing for routine allocation or implementation choices. Continue independent authorized work when a finding is blocked or requires approval.

1. Discover Repository Tooling

1. Work from the repository root and read the applicable repository instructions.
2. Inspect the worktree before editing. Preserve user-owned changes and identify files already modified by others.
3. Inspect package scripts, their wrappers, workspace configuration, and diagnostic configuration files to identify the project's lint, typecheck, and other configured diagnostic commands. Use installed dependencies as supporting evidence; a package's presence alone does not establish that the repository uses it for diagnostics.
4. Prefer repository-defined commands and preserve their configuration and wrapper behavior. Use bun run <script> for package scripts and bunx <tool> for local executables by default, unless explicit repository instructions require another runner. Identify separate TypeScript projects and diagnostic commands hidden behind aggregates that stop after the first failure.
5. For React or React Native projects, include React Doctor automatically unless the user's narrower scope excludes it. Prefer a repository-defined script or installed version; otherwise use bunx react-doctor@latest. Record the resolved version when available.
6. Use the requested tools, files, rules, or findings to limit the diagnostic scope. When no narrower scope is given, select the applicable project-wide diagnostics discovered above. Identify baseline commands and any supported scoped check or fix commands before running them; inspect scripts to distinguish non-mutating checks from fix operations.

Complete this phase when every applicable diagnostic source has identified commands and configurations or an exact recorded discovery blocker.

2. Collect the Diagnostic Baseline