The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and executes the react-doctor package from the NPM registry using npx react-doctor@latest. It also fetches official fix recipes and documentation from the vendor's domain at https://www.react.doctor/prompts/rules/.
[COMMAND_EXECUTION]: The skill performs shell command execution to facilitate the audit process. This includes running npx for structured JSON reports and git rev-parse to tag generated plans with the current commit hash.
[PROMPT_INJECTION]: The skill includes robust defensive instructions to handle potential prompt injections hidden in analyzed source code. The instructions explicitly command the agent to treat repository content as data rather than instructions and to ignore phrases like "ignore previous instructions" if encountered in file contents.
Ingestion points: The skill ingests the entire contents of a React codebase for analysis.
Boundary markers: Hard Rule 4 in SKILL.md serves as a clear boundary, defining all repository content as inert data.
Capability inventory: The skill has the capability to execute shell commands (npx, git) and write files to the plans/ directory.
Sanitization: The skill instructs the agent to flag any detected injection attempts as audit findings rather than obeying them, preventing the data from influencing the agent's control flow.

improve-react