track-roadmap
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill represents a legitimate productivity tool for roadmap management. It provides clear structural requirements for the ROADMAP.md file and implements a safe interaction model where the user must approve all proposed updates.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it reads and processes data from the project's codebase during its audit and discovery modes.\n
- Ingestion points: The skill scans the ROADMAP.md file and project source files (including comments and TODOs) to evaluate feature implementation status, as detailed in reference/MODES.md.\n
- Boundary markers: The main skill instructions in SKILL.md explicitly mandate that the agent must "propose a change, confirm with the user, write the file." This requirement for human confirmation effectively prevents embedded instructions in code from triggering unauthorized actions without user knowledge.\n
- Capability inventory: The skill is capable of reading project files, writing to ROADMAP.md, and invoking the track-session skill.\n
- Sanitization: While no specific input sanitization or filtering logic is mentioned for the codebase scan, the human-in-the-loop validation process acts as a sufficient security control.
Audit Metadata