The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches multiple instructional and configuration files from the gh-aw repository on GitHub. These files are used to guide the agent's behavior for specialized tasks such as workflow design, debugging, and optimization.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and follows instructions from external and local sources without sanitization.
Ingestion points: Files are loaded from the remote github/gh-aw repository and the local .github/aw/instructions.md file.
Boundary markers: The skill does not implement explicit delimiters or isolation instructions when loading these external prompts; it is instructed to "follow it directly."
Capability inventory: The agent is granted the capability to design, create, update, and debug GitHub workflows, as well as invoke telemetry-related skills.
Sanitization: No validation, escaping, or filtering of the loaded content is performed before it is integrated into the agent's execution context.

agentic-workflows