The Agent Skills Directory

[SAFE]: The skill is a planning utility that does not perform network operations, file system modifications, or command execution itself. It provides structured guidance for a 'main Agent' to coordinate subagents.
[INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting user-provided coding tasks and transforming them into subagent instructions. While this creates a potential surface for indirect injection (where a malicious task description influences subagent behavior), the skill mitigates this by enforcing 'The Iron Law' of independent inputs/outputs and requiring explicit 'output contracts' for every dispatched slice. Capability inventory includes the platform's Agent tool with recommended isolation: "worktree" security controls.

parallel-implementation