validate-merged
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill evaluates repository history and recent code merges which represent an external, untrusted data ingestion point. Ingestion points occur during the parsing of merge frontiers and commit bundles via project roots. Boundary markers are absent within the instructions to isolate raw code data from natural language commands. Capability inventory includes executing local shell operations via Bash and dispatching downstream finder and refuter Tasks. Sanitization is not specified to filter or neutralize potentially malicious instructions embedded inside code diffs or commit messages.
Audit Metadata