The Agent Skills Directory

[Indirect Prompt Injection Considerations]: The skill sequences tasks by ingesting caller-supplied research and plan inputs. Processing external data is an ingestion point that requires consideration of how instructions within that data might interact with the agent's logic.
Ingestion points: The skill receives data through the evidence and task inputs as defined in SKILL.md.
Boundary markers: No specific delimiters or markers are documented to encapsulate the ingested evidence to prevent instruction leakage.
Capability inventory: The skill coordinates the creation of artifacts in the .copilot-tracking/ directory and manages the execution flow between implementation and review phases.
Sanitization: The instructions do not specify validation or sanitization logic for the incoming external data.

rpi-quick