agentic-workflows

Pass

Audited by Gen Agent Trust Hub on Jul 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill is configured to fetch numerous instructional and prompt files from the 'github/gh-aw' repository when specific tasks are requested. These files are not present in the local workspace.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by design.
  • Ingestion points: It checks for and loads instructions from '.github/aw/instructions.md' within the local repository.
  • Boundary markers: No explicit boundary markers or isolation instructions are used when loading these local instructions.
  • Capability inventory: The skill allows the loaded instructions to override default behaviors and directs the agent to perform workflow design, creation, and debugging tasks based on the content.
  • Sanitization: The skill does not perform validation or sanitization of the local instructions file, explicitly stating that repository overlay instructions override upstream defaults.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 19, 2026, 06:46 AM
Security Audit — agent-trust-hub — agentic-workflows