The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as nub run lint, nubx tsc, and bunx eslint to perform diagnostic checks and apply code fixes. These are standard development operations.
[EXTERNAL_DOWNLOADS]: The skill fetches the latest version of the react-doctor package during execution via the command nubx react-doctor@latest. This is a remote download of a well-known development tool.
[PROMPT_INJECTION]: The skill processes untrusted data from diagnostic tool outputs, creating an indirect prompt injection surface. It documents ingestion points in SKILL.md Phases 1 and 2, and mitigates risks through a mandatory verification phase (Phase 2) where the agent must confirm findings in the source code before using its file-writing capabilities in Phase 3 or further subprocess commands in Phase 4. The skill also enforces boundary markers by requiring minimal changes and adherence to repository conventions.

fix-repository-diagnostics