interrogate
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code and diffs within a sub-agent prompt template, creating an attack surface for indirect prompt injection. Ingestion points: Untrusted data enters the agent context via the {DIFF_OR_FILES} variable in the 'references/reviewer-prompt.md' file. Boundary markers: The template uses standard Markdown headers (e.g., '## Code Under Review') to delineate sections, which lacks the security of unique delimiters or XML-based tagging to prevent sub-agents from being influenced by instructions embedded in the code. Capability inventory: Worker agents are restricted to 'read-only' operations and are prohibited from editing files, executing destructive commands, or calling external systems, as specified in 'SKILL.md' and 'references/interrogate-panel.md'. Sanitization: The skill instructions do not specify any validation, escaping, or filtering of the code content before it is interpolated into the prompts.
Audit Metadata