The Agent Skills Directory

[COMMAND_EXECUTION]: Employs ripgrep (rg) and xargs to search for specific code patterns and hook usages across the repository.
[COMMAND_EXECUTION]: Uses pnpm exec to run local development tools like agent-react-devtools for profiling re-renders.
[DYNAMIC_EXECUTION]: Includes a Node.js script utilizing the playwright library to automate browser interactions for counting RxJS subscriptions at runtime.
[EXTERNAL_DOWNLOADS]: Provides instructions to download and extract a specific version of a library using npm pack react-rx@6.0.1 for comparative testing.
[DATA_EXFILTRATION]: Accesses the STUDIO_AUTH_TOKEN environment variable to authenticate the automated browser session. The token is handled locally and passed to a localhost URL via a hash fragment.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates an audit process that reads local source code, creating an ingestion surface for potential instructions within the audited data.
Ingestion points: Search tool output and local file content (e.g., SKILL.md mentions useValuePreview.ts).
Boundary markers: None specified for the code analysis task.
Capability inventory: Shell command execution, browser automation, and file system modification (replacing node_modules content).
Sanitization: Uses encodeURIComponent for local URL construction, but does not provide sanitization for the audited code content.

react-rx-hook-audit