The Agent Skills Directory

[PROMPT_INJECTION]: Auditing external or AI-generated code creates an indirect prompt injection surface where malicious code could try to influence the agent.
Ingestion points: Analyzes code as defined in SKILL.md.
Boundary markers: Uses 'Contract Integrity' and 'Security Sanitization' rules.
Capability inventory: Authorized to halt commits, run tests, and remediate code.
Sanitization: Directs use of Zod and Snyk AST.
[COMMAND_EXECUTION]: Instructions include running standard tools like bun x tsc, eslint, and prettier.
[EXTERNAL_DOWNLOADS]: Uses bun x and web_search for standard development tasks involving trusted services.

strict-auditor