team-planner
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill orchestrates multiple agents where the output of sub-agents (stored in the 'assignments' SQL table) is subsequently consumed by a 'synthesizer' agent in Phase 6.
- Ingestion points: Phase 6 (Synthesizer) reads data from the 'team' and 'assignments' SQL tables, as well as referenced artifacts produced by sub-agents (SKILL.md).
- Boundary markers: None identified; the synthesizer is instructed to read the tables and artifacts directly without explicit delimiters or warnings to ignore embedded instructions.
- Capability inventory: The skill utilizes the 'task' tool (for agent spawning), 'sql' tool (for state management), 'read_agent', and 'write_agent' tools (for session interaction) (SKILL.md).
- Sanitization: No evidence of sanitization or validation of the sub-agent outputs before they are processed by the synthesis agent.
Audit Metadata