claude-devfleet
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs the user to manually install the DevFleet server from a third-party GitHub repository (github.com/LEC-AI/claude-devfleet) during the setup process.
- [INDIRECT_PROMPT_INJECTION]: The skill acts as an orchestrator for sub-agents, creating an attack surface where instructions could be injected into mission prompts.
- Ingestion points: The
plan_projectandcreate_missiontools take prompt strings as input, which are used to direct the activity of sub-agents. - Boundary markers: The skill does not define specific delimiters or protective instructions to prevent sub-agents from following commands embedded in these prompts.
- Capability inventory: Sub-agents are described as having full tooling and operating within git worktrees, which implies capabilities for command execution and filesystem interaction.
- Sanitization: No sanitization, escaping, or validation of the prompts is mentioned in the skill instructions.
- [NO_CODE]: This skill consists entirely of a SKILL.md file providing instructions and tool descriptions, with no external scripts, binaries, or source code files included.
Audit Metadata