triage
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script using
uv run python3 $CLAUDE_PROJECT_DIR/tools/brain.py. It passes arguments such as<target>,<finding>, and<kill reason>which are extracted directly from processed data files. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from
findings.mdorfindings.json(Ingestion point). This data is used to drive the triage logic and populate shell command arguments. - Boundary markers: None identified. The instructions do not specify using delimiters or 'ignore embedded instructions' blocks when reading the finding files.
- Capability inventory: Executes subprocess calls via
uv runand launches sub-agents (validator). - Sanitization: No explicit sanitization or escaping of characters is mentioned before interpolating finding text into the shell command execution step.
Audit Metadata