The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a command-line utility called taskmd to query task details and update their statuses.
[INDIRECT_PROMPT_INJECTION]: The skill reads task descriptions and acceptance criteria from local files to guide its planning and subagent creation.
Ingestion points: Data is ingested from the tasks/ directory via the Read tool and taskmd output.
Boundary markers: The instructions do not define delimiters to separate untrusted task content from the agent's core logic.
Capability inventory: The agent has access to Bash for command execution, Write/Edit for file system changes, and the Agent tool to delegate tasks to subagents.
Sanitization: No sanitization or safety checks are performed on the ingested task descriptions before they are processed by the planning logic.

divide-and-conquer