The Agent Skills Directory

[PROMPT_INJECTION]: The skill's orchestration workflow creates an indirect prompt injection surface by passing user-supplied instructions to sub-agents.
Ingestion points: User prompts provided to the plan_project and create_mission tools in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions to sub-agents to disregard instructions embedded within mission prompts.
Capability inventory: Spawned agents are described as having 'full tooling' access within isolated git worktrees.
Sanitization: No validation or sanitization of the input prompt parameters is documented.
[NO_CODE]: The skill consists entirely of markdown documentation and tool definitions with no included scripts, binaries, or dynamic code execution logic.

claude-devfleet