The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script using the uv tool (e.g., uv run scripts/markdown_to_task_pool.py tasks.md). This is part of the core functionality for converting markdown files into structured task data.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes untrusted data from user-provided markdown files to generate task descriptions and worker instructions. Malicious instructions embedded in a checklist item could potentially influence the behavior of spawned worker agents.
Ingestion points: The SKILL.md specifies reading from user-supplied markdown files like todo.md or checklist.md via the markdown_to_task_pool.py script.
Boundary markers: None are present in the provided instructions or the Python parsing snippet to isolate or disregard instructions within the checkbox items.
Capability inventory: The skill uses TeamCreate, TaskCreate, and Agent (which spawns sub-agents) in addition to shell execution via uv run.
Sanitization: There is no evidence of input validation or sanitization applied to the text extracted from the markdown checklist before it is used in task creation and agent prompts.

swarm-from-markdown