claude-devfleet
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions direct the user to download and install an external server from a non-whitelisted GitHub repository (https://github.com/LEC-AI/claude-devfleet) and connect it via MCP.
- [INDIRECT_PROMPT_INJECTION]: The multi-agent orchestration pattern creates a vulnerability surface where the orchestrator agent ingests and acts upon data generated by sub-agents.
- Ingestion points: Data returned by the
get_reporttool (referenced in SKILL.md). - Boundary markers: None specified; the skill does not include delimiters or instructions for the orchestrator to ignore commands embedded in sub-agent reports.
- Capability inventory: The orchestrator has the capability to create git worktrees, merge code, and dispatch further missions based on project status.
- Sanitization: No sanitization or validation of the sub-agent report content is described in the workflow.
Audit Metadata