agentic-workflows
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to load and prioritize instructional content from the local repository being processed, specifically targeting
.github/aw/instructions.md. - Ingestion points: The agent is instructed to load
.github/aw/instructions.mdfrom the user's workspace if it exists. - Boundary markers: There are no boundary markers or instructions to treat the loaded content as untrusted; the agent is explicitly told to "follow it directly" and that these instructions "override upstream defaults."
- Capability inventory: The skill facilitates designing, creating, and debugging workflows, which involves file system manipulation and potential tool execution based on the loaded instructions.
- Sanitization: No sanitization or validation logic is defined for the content loaded from the project workspace.
- [EXTERNAL_DOWNLOADS]: The skill retrieves a large collection of instruction frameworks from the official GitHub organization repository (
github/gh-aw). This is the intended core behavior of the skill for fetching workflow logic.
Audit Metadata