claude-devfleet
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it ingests untrusted text via the prompt parameter in the plan_project and create_mission tools. This content is then processed by sub-agents that possess full tooling capabilities within isolated git worktrees.\n
- Ingestion points: prompt parameters in plan_project and create_mission tools defined in SKILL.md.\n
- Boundary markers: None identified in the skill instructions.\n
- Capability inventory: Sub-agents operate with full tool access including file system writes and command execution in isolated worktrees.\n
- Sanitization: No validation or sanitization of the injected prompts is described in the skill.
Audit Metadata