rpi-implement
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Instruction Processing: The skill processes external implementation plans and details stored in the .copilot-tracking/ directory to guide its actions. This is a primary function of the agent.\n
- Ingestion points: Implementation plans and details files (e.g., in SKILL.md step 1 and references/implementation.md section 1).\n
- Boundary markers: The instructions do not specify explicit delimiters or 'ignore' warnings for the content of these ingested files.\n
- Capability inventory: The skill can modify files, execute subagents via runSubagent, and run validation commands.\n
- Sanitization: No specific sanitization or validation of the plan content is described before execution.\n- Automated Validation Execution: As part of its workflow, the skill identifies and executes validation commands (such as npm run scripts) defined within the implementation plans. This automation is intended to confirm the correctness of implemented changes.\n- Information Leakage Prevention: The skill includes explicit constraints to prevent internal project metadata, such as file paths from the .copilot-tracking/ directory, from appearing in production code, comments, or commit messages. This practice helps maintain the security and cleanliness of the shipped codebase.
Audit Metadata