The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill directs the agent to read and analyze external source code and configuration files. This creates a surface where malicious instructions embedded in the audited files (e.g., in code comments) could attempt to manipulate the agent's report or bypass specific security checks.
Ingestion points: app/, components/, contracts/, packages/nextjs/scaffold.config.ts, packages/nextjs/styles/globals.css, and packages/nextjs/contracts/externalContracts.ts.
Boundary markers: Absent. The instructions do not provide specific delimiters or ignore-embedded-instruction warnings for the audited content.
Capability inventory: The agent is instructed to use tools such as grep, vercel, yarn, and forge to audit the project environment and code.
Sanitization: None. The files are processed as raw text without sanitization or escaping.
[COMMAND_EXECUTION]: The skill provides the agent with several shell commands to facilitate the audit process. While these are used for development and QA purposes, they involve interacting with the system environment and deployment tools.
Evidence: The skill suggests executing grep -rn, vercel env ls, yarn verify, and forge verify-contract to search for patterns, check environment variables, and verify smart contracts.

qa