The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to ingest and act upon external, potentially untrusted data sources which could contain malicious instructions. This occurs when the agent inspects file diffs, reviews task criteria, and analyzes logs from builds, tests, or linters. Because the agent is also instructed to execute shell commands based on these inputs without explicit sanitization or boundary markers, a specially crafted diff or log file could attempt to influence the agent's behavior. 1. Ingestion points: The agent processes task completion criteria, file diffs, and output from build, test, and linter processes (SKILL.md). 2. Boundary markers: The instructions do not define specific delimiters or security warnings to help the agent distinguish between its instructions and the data being reviewed. 3. Capability inventory: The skill directs the agent to perform file system reads (diff review), write to the workspace (VALIDATION.md), and execute shell commands (running builds, compilers, and tests). 4. Sanitization: There are no specific steps mentioned to sanitize or validate the content of external files or logs before processing.

check-work