express-intent-in-code

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of code snippets or files which serves as an ingestion point for potential malicious instructions.
  • Ingestion points: User-supplied code provided to the agent for refactoring (SKILL.md).
  • Boundary markers: The instructions lack specific delimiters or directions to disregard embedded instructions (e.g., in comments or string literals) within the code being processed.
  • Capability inventory: The skill has the capability to write files (verified edits) and execute shell commands (running lint and behavior checks).
  • Sanitization: There is no mention of sanitizing or validating the input code before the agent processes it or executes associated test suites.
  • [COMMAND_EXECUTION]: The skill instructions require the agent to execute shell-based verification tools and scripts found within the local environment.
  • Evidence: "After verified edits, run fresh lint and the supplied behavior checks..." (SKILL.md).
  • Risk: The skill relies on "supplied behavior checks," which are arbitrary scripts or executables. Executing these tools poses a risk if the repository containing the code and tests is untrusted or contains malicious scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 07:13 PM
Security Audit — agent-trust-hub — express-intent-in-code