The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external orchestration data from plan.md and manifest.json to drive agent behavior.
Ingestion points: The skill reads structured data from the .claude/plans/ directory to determine task dependencies and agent waves.
Boundary markers: The skill encourages the use of external briefing files for spawned agents to isolate task context and minimize cross-task instruction leakage.
Capability inventory: The skill has the capability to spawn multiple specialized agents and perform file system deletions.
Sanitization: The instructions direct the agent to follow the plan manifest directly without explicit validation of the manifest's contents or the plan-id path components.
[COMMAND_EXECUTION]: The skill uses standard shell commands to manage its working directory environment.
Evidence: The skill executes rm -rf .claude/plans/<plan-id>/ to clean up temporary orchestration data after a task project is completed. This is a functional behavior intended for environment resource management.

coordinator