The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a bundled Node.js script to execute local development tools such as ESLint, the TypeScript compiler (tsc), and Git for repository analysis.
Evidence: scripts/collect-quality-evidence.mjs invokes spawnSync to run these binaries from the project's node_modules/.bin directory.
Safeguards: The script implements strict path validation using realpathSync and insideRoot checks to prevent the audit from accessing files outside the repository boundary. Commands are executed with a --no-fix or --noEmit flag to ensure they are read-only.
[INDIRECT_PROMPT_INJECTION]: As an auditing tool, the skill is designed to ingest and process untrusted source code and configuration files from the target repository, which creates a potential surface for indirect prompt injection.
Ingestion points: The agent is instructed to read package.json, nest-cli.json, and all source files within the audit scope to identify architectural and design issues.
Capability inventory: The skill's environment is restricted to read-only file system access; it cannot modify the repository, access the network, or perform privilege escalation.
Evidence Chain: The skill aggregates untrusted code into a structured report but relies on the agent's internal reasoning to differentiate between code evidence and potential malicious instructions embedded in comments or string literals.

nestjs-code-audit