grill-me
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes untrusted data, such as project plans and codebase files, which creates a potential surface for indirect prompt injection where adversarial instructions could attempt to influence the agent.
- Ingestion points: Processes user-provided plans, PRDs, and repo documentation as described in
SKILL.md(Step 0). - Boundary markers: No specific delimiters or 'ignore embedded instructions' markers are defined for the ingested data.
- Capability inventory: Utilizes search MCP tools (
githubSearchCode), launches background subagents for research, and writes report files to the filesystem (Step 5). - Sanitization: Input data is not explicitly sanitized or validated before processing.
- [SAFE]: No malicious patterns, exfiltration, or unauthorized execution were detected. The skill's file access and persistence behaviors are appropriate for its purpose as a document reviewer.
Audit Metadata