The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches the nvm installation script from its official repository on GitHub to align the project's Node.js environment.
[REMOTE_CODE_EXECUTION]: Executes the downloaded nvm script using a piped shell command (curl | bash). This setup targets a well-known, reputable service for developer tooling.
[COMMAND_EXECUTION]: Performs automated package installation and project building using npm install, yarn install, pnpm install, npm run build, and npm test across the project structure.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it executes installation and build scripts directly from the audited project's repository. Malicious code within the target project's configuration or tests could trigger unintended command execution during the audit process.
Ingestion points: Reads and processes package.json, configuration files, and source code from the external project being audited.
Boundary markers: No specific boundary markers or warnings are used when executing scripts defined in the project's configuration files.
Capability inventory: The skill uses powerful tools including Bash, Agent, Write, Edit, and Read across a multi-agent orchestration wave.
Sanitization: The skill does not implement explicit validation or sanitization of project-provided scripts (e.g., in package.json) before they are executed.

react-health-audit