The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is configured to fetch a comprehensive set of instruction files (markdown) from a remote repository (github/gh-aw). These files provide the core logic for designing, creating, and debugging workflows.
[INDIRECT_PROMPT_INJECTION]: The skill operates as a dynamic instruction loader, ingesting content from both a remote repository and a local project-specific file (.github/aw/instructions.md). This creates a surface for potential instruction injection if these sources are compromised.
Ingestion points: Remote files within the github/gh-aw repository and the local file .github/aw/instructions.md.
Boundary markers: The skill establishes a precedence rule (local overrides remote) but does not implement specific delimiters or safety warnings to mitigate the risk of malicious instructions within the loaded content.
Capability inventory: The skill's primary function is routing and context loading; it facilitates the agent's transition to specialized task-handling states based on file content.
Sanitization: No explicit sanitization or filtering of the fetched markdown content is performed before integration into the agent's active context.

agentic-workflows