The Agent Skills Directory

Data Ingestion from Workspace Files: The skill reads implementation plans and task definitions from the .copilot-tracking/ directory. This is a common pattern for development-oriented agents that rely on project context to perform their tasks.
Ingestion points: Plan files in .copilot-tracking/plans/ and reference files such as references/implementation.md (identified in SKILL.md and references/implementation.md).
Boundary markers: The instructions do not specify explicit delimiters or "ignore" instructions to separate plan content from the agent's logic.
Capability inventory: The skill possesses capabilities for writing files to the workspace, delegating tasks to subagents, and prompting the user for input.
Sanitization: The skill does not include specific sanitization or filtering steps for the ingested plan data.
Subagent and Skill Delegation: The skill identifies and delegates work phases to other available subagents or skills based on their names and descriptions. It establishes specific prompts and constraints for these delegated workers, which helps maintain consistency and control over the implementation scope.
Local State Tracking: Implementation evidence and progress markers are maintained in Markdown files within the workspace. This approach provides a transparent audit trail of changes and allows for persistent state management across different execution cycles.

rpi-implement