learn-decompose
Pass
Audited by Gen Agent Trust Hub on Jul 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash and Glob tools to identify target files and interact with the maestro CLI tool. This usage is aligned with the skill's primary function of analyzing code patterns within a development environment.
- [PROMPT_INJECTION]: The skill processes external source code, which introduces an indirect prompt injection surface where malicious code content could attempt to influence the agent's logic. This risk is effectively mitigated by the skill's enforcement of a specific output directory (.workflow/) and a human-in-the-loop preview step using the request_user_input tool.
- Ingestion points: Local source files identified in the target path (Phase 1).
- Boundary markers: None present for the content of the analyzed files.
- Capability inventory: Tools include Bash, Write, Edit, and spawn_agents_on_csv (Phase 2 and Phase 4).
- Sanitization: Mitigated by the mandatory preview gate requiring user confirmation for all writes.
Audit Metadata