The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a repository overlay mechanism that automatically loads and prioritizes instructions from a local file, creating a surface for workspace-based instruction override.
Ingestion points: The skill reads content from .github/aw/instructions.md in the user's workspace to override default behaviors.
Boundary markers: There are no explicit delimiters or "ignore instructions" markers used when interpolating the overlay content into the agent's context.
Capability inventory: While the router itself is simple, it dispatches to complex prompt files (e.g., designer.md, debug-agentic-workflow.md) that likely invoke file system and terminal tools.
Sanitization: The skill lacks validation or sanitization for the local instructions it ingests.
[EXTERNAL_DOWNLOADS]: The skill is designed to fetch a wide array of prompt and configuration files from the github/gh-aw repository. These remote resources are central to the skill's functionality, providing the specific logic for designing, debugging, and managing agentic workflows. These downloads target the vendor's own official repository infrastructure.

agentic-workflows