triage

Warn

Audited by Gen Agent Trust Hub on Jul 23, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to verify external pull requests by checking them out and running "relevant tests or commands". This behavior introduces a risk of executing malicious code embedded within an untrusted PR's test suite or build scripts.
  • Evidence: SKILL.md
  • "Verify the claim. ... For a PR, confirm the diff does what it claims — check it out, run the relevant tests or commands."
  • [COMMAND_EXECUTION]: The skill uses shell commands to perform repository operations and verify external contributions, which can be manipulated if the source is untrusted.
  • Evidence: SKILL.md
  • "check it out, run the relevant tests or commands."
  • [PROMPT_INJECTION]: The skill processes untrusted input from external contributors (issue and PR bodies, comments, and diffs) and possesses the capability to execute shell commands, creating a surface for indirect prompt injection.
  • Ingestion points: The skill reads issue and PR bodies, comments, author information, and diffs (SKILL.md).
  • Boundary markers: Absent. There are no instructions to ignore or treat embedded instructions as data.
  • Capability inventory: The skill is authorized to check out PR code and execute "relevant tests or commands" (SKILL.md).
  • Sanitization: Absent. The skill does not describe any validation or filtering of the PR content before execution.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 23, 2026, 07:33 AM
Security Audit — agent-trust-hub — triage