todo-harvest
Pass
Audited by Gen Agent Trust Hub on Jun 27, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external codebase content.
- Ingestion points: The
scripts/harvest.pytool (referenced inSKILL.md) extracts data fromTODO,FIXME, andHACKmarkers found in local source files. - Boundary markers: Absent. There are no instructions to wrap or delimit the extracted comment text to prevent the model from following instructions embedded within the comments.
- Capability inventory: The skill utilizes the
Bashtool to execute a Python script and Git commands (git ls-files,git blame). - Sanitization: The instructions do not implement any sanitization or filtering of the comment content before it is processed by the AI.
Audit Metadata