The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches the nvm installation script from the official GitHub repository for the nvm-sh project during the mandatory environment setup phase.
[REMOTE_CODE_EXECUTION]: Executes the downloaded nvm shell script directly via bash. This is a standard installation pattern for this developer tool and originates from a well-known, trusted service.
[COMMAND_EXECUTION]: Performs several potentially high-impact shell operations including global installation of package managers (yarn, pnpm), local dependency installation (npm install), and the execution of project-defined test and build scripts. These actions are required to generate the audit's technical metrics.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the repository being audited (source code, configuration files, and documentation). While the skill possesses extensive capabilities (shell access, file modification, and sub-agent spawning), it does not specify explicit sanitization or boundary markers to isolate the agent's reasoning from instructions that might be embedded in the audited code.
Ingestion points: Project-level files read and scanned by the analyzer agents (e.g., package.json, *.ts, README.md).
Boundary markers: Absent.
Capability inventory: Shell command execution (Bash), file writing (Write), and sub-agent delegation (Agent).
Sanitization: No specific filtering or escaping of project content is mentioned before processing.

nestjs-health-audit