planning-and-task-breakdown
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external and potentially untrusted data to generate implementation plans, which could be manipulated by malicious content within those sources.
- Ingestion points: The skill instructs the agent to read project specifications and relevant codebase sections (SKILL.md, Step 1).
- Boundary markers: No explicit boundary markers or instructions to ignore embedded commands within ingested data are defined.
- Capability inventory: The skill includes instructions to write plan and task documents to the local file system (
tasks/plan.mdandtasks/todo.md). - Sanitization: The skill lacks explicit sanitization or filtering of external content before it is interpolated into the generated output files.
- Mitigation: The skill emphasizes a "read-only mode" during planning and mandates human review and approval of the plan before implementation begins, providing a strong manual control against automated exploitation.
Audit Metadata