context-audit
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from context files (CLAUDE.md, MEMORY.md, etc.) and processes it for analysis.
- Ingestion points: The skill reads multiple files that may contain user-provided or external content, specifically
CLAUDE.md,AGENTS.md,SOUL.md,USER.md,ACCESS_POLICY.md,HEARTBEAT.md, andMEMORY.md(Procedure Step 1 and 2). - Boundary markers: The skill does not implement explicit delimiters or 'ignore' instructions when passing the aggregated content to the evaluation runner (
gbrain eval cross-modalin Step 4). - Capability inventory: The skill has access to
shellandreadtools, though it is explicitly configured asmutating: false. - Sanitization: There is no evidence of sanitization or filtering of the file content before it is processed by the agent or the evaluation judge.
Audit Metadata