skills/gannonh/skills/ps-interrogate/Gen Agent Trust Hub

ps-interrogate

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses git diff main...HEAD in SKILL.md to identify the code changes requiring review. This is a standard use of version control tools within a developer-oriented skill.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from feature branches via git diffs, creating a surface for indirect prompt injection where malicious comments in the code could attempt to influence the reviewer models. 1. Ingestion points: The output of git diff is read in SKILL.md (Step 1) and passed to subagents in references/reviewer-prompt.md via the {DIFF_OR_FILES} placeholder. 2. Boundary markers: The prompt template in references/reviewer-prompt.md lacks explicit structural delimiters or 'ignore' instructions around the untrusted diff content, although it is contained within a labeled section. 3. Capability inventory: The skill's capabilities are restricted to reading files, spawning subagent reviewers, and synthesizing text findings. The skill explicitly directs the agent 'Do NOT auto-apply changes', which serves as a critical safety barrier. 4. Sanitization: The skill does not perform sanitization or filtering of the diff content before it is analyzed by the subagents.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 01:26 AM
Security Audit — agent-trust-hub — ps-interrogate