The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses ls -La and grep commands via Bash to locate task-specific directories within .humanlayer/tasks/. This is used for navigational purposes and to handle symlinks correctly.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data including plan documents, tickets, and task artifacts.
Ingestion points: Reads files from .humanlayer/tasks/ such as ticket.md, task.md, research documents, and design discussions.
Boundary markers: None explicitly defined in the prompt instructions for external data ingestion.
Capability inventory: Access to Read, Write, Edit, MultiEdit tools, as well as Bash for directory listing and the Agent tool for sub-agent delegation.
Sanitization: No explicit sanitization or escaping of the external document content is mentioned before it enters the agent context.
[SAFE]: The referenced template file references/iterate_implementation_final_answer.md contains static markdown and placeholder text for standardized reporting, which does not present a security risk.

iterate-implementation