The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection vulnerability surface because it processes untrusted user input to generate instructions for sub-agents.
Ingestion points: The prompt parameter in the plan_project and create_mission tools as described in SKILL.md.
Boundary markers: Absent; there are no instructions to use delimiters or warnings for the agent to ignore instructions embedded in the user data.
Capability inventory: Orchestrated agents are granted 'full tooling' and the ability to perform file modifications and merges in git worktrees.
Sanitization: No evidence of sanitization or validation of the input prompt is provided in the skill instructions.
[COMMAND_EXECUTION]: The documentation includes a shell command for configuring the required MCP server.
Evidence: claude mcp add devfleet --transport http http://localhost:18801/mcp found in SKILL.md.

claude-devfleet