recipe-generalization-sweep
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on executing local Python scripts (e.g., python3 .plan/execute-script.py) to perform its core functions, including configuration management, architecture enumeration, and writing solution outlines. These commands target internal project scripts designed for the plan-marshall ecosystem.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its core workflow involves reading the full content of untrusted skill files and scripts. If a processed file contains malicious instructions, it could potentially influence the agent's adjudication logic. 1. Ingestion points: Full-read of all files within the plan-marshall bundle, specifically SKILL.md, standards/, references/, and scripts/ files. 2. Boundary markers: The instructions do not define clear delimiters or use 'ignore instructions' guards when processing the content of the audited files. 3. Capability inventory: The agent has access to Bash, Write, and python3 execution capabilities, which could be abused if the analysis phase is compromised. 4. Sanitization: Verbatim quotes from audited files are recorded and written to solution_outline.md without apparent sanitization.
Audit Metadata