The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the workmux CLI to manage agent sessions and execute shell commands within specific git worktrees (e.g., workmux run agent-a -- pytest). This behavior is the primary function of the coordinator skill.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user instructions and external markdown files to generate prompt files for sub-agents.
Ingestion points: User tasks, conversation history, and markdown files (plan/spec) read via the Read tool.
Boundary markers: None explicitly defined in the instructions for isolating external content within the generated prompt files.
Capability inventory: The skill has access to the Bash tool and can execute arbitrary shell commands via workmux run.
Sanitization: No explicit sanitization or escaping of ingested data is described before interpolation into prompt files.
[DYNAMIC_EXECUTION]: The skill dynamically creates temporary markdown files using mktemp and cat to store instructions, which are then loaded and executed by new agent sessions using workmux add -P <file>. This is a core component of the agent dispatching workflow.

coordinator