The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several command-line tools to verify security posture. This includes npm audit or pnpm audit for dependency checks, npx gitleaks for secret detection, and grep for scanning client bundles and environment variables. It also uses curl to perform 'outside-in' checks on protected endpoints to verify HTTP status codes and response sizes.
[EXTERNAL_DOWNLOADS]: The skill facilitates the use of external security tools and plugins. It invokes npx gitleaks, which may download the package at runtime, and mentions rls-audit for Supabase Row Level Security checks. Additionally, it suggests installing two plugins from the same author (MartinOlivero/proof-first and MartinOlivero/saas-legal-docs) to extend its auditing capabilities.
[INDIRECT_PROMPT_INJECTION]: As a security auditing tool, the skill naturally ingests untrusted data which creates a surface for indirect prompt injection.
Ingestion points: The skill processes the project's entire codebase, git history, environment variable files, and the output of HTTP requests to live endpoints.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat the audited code as data rather than instructions, which could allow malicious code to influence agent behavior.
Capability inventory: The skill has access to shell execution (curl, npm, grep), file system reading (code, .env, git history), and the ability to recommend plugin installations.
Sanitization: The instructions do not define any sanitization or validation steps for the content being audited before it is processed by the agent.

pre-ship-security