The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and analyzing the source code of external full-stack applications. This creates an inherent surface for indirect prompt injection where malicious instructions hidden in the application's comments or logic could attempt to influence the agent's behavior during the audit or fix phases.
Ingestion points: The agent reads the frontend, backend, and configuration layers of the application being audited (SKILL.md, Execution process section).
Boundary markers: The execution process does not specify the use of clear delimiters or instructions for the agent to ignore potential directives embedded within the audited code.
Capability inventory: The skill's instructions involve file system modifications to apply security fixes, execution of build scripts (npm run build), version control operations (git commit/push), and network requests to the Vercel API via curl commands.
Sanitization: There is no defined process for sanitizing or filtering the ingested source code before the agent processes it for auditing.
[COMMAND_EXECUTION]: The skill guides the user or agent through executing standard development and deployment commands. This includes build verification using npm run build, version control operations using git, and managing environment variables through the Vercel API using curl. These commands target legitimate development tools and a well-known cloud service provider.

vibe-proof