The Agent Skills Directory

Remote Instruction Fetching: The skill is designed to load workflow instructions and prompt templates from the github/gh-aw repository. This mechanism centralizes workflow logic by fetching configuration files (such as .github/aw/designer.md and .github/aw/patterns.md) from an external source to guide the agent's behavior.
Indirect Prompt Injection Surface: The skill facilitates the loading of instructions from the local repository via the .github/aw/instructions.md file, which is prioritized over default settings. This creates a surface where data within the repository can influence the agent's core instructions.
Ingestion points: The agent is instructed to read .github/aw/instructions.md from the current repository if it exists.
Boundary markers: There are no explicit delimiters or warnings provided to the agent to distinguish these repository-level instructions from its system-level safety guidelines.
Capability inventory: While the router skill itself does not perform file writes or network operations, it directs the agent to load other skill files (like skills/otel-queries/SKILL.md) which may possess broader capabilities.
Sanitization: The skill does not implement validation or filtering for the content loaded from the local repository overlay.

agentic-workflows