agentic-workflows

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFE
Full Analysis
  • Prompt Routing Architecture: The skill is designed to identify user tasks and load corresponding instruction files from the .github/aw/ and .github/skills/ directories. This is a modular approach to managing complex workflows within the repository.
  • Instruction Ingestion Pattern: The agent is instructed to read and follow instructions from various markdown files. While this creates an ingestion point where external file content influences agent behavior, this is a core functional requirement of a router skill.
  • Ingestion points: Multiple markdown files located in .github/aw/ and .github/skills/ (e.g., create-agentic-workflow.md, debug-agentic-workflow.md).
  • Boundary markers: Instructions are followed directly after loading; the skill does not use explicit delimiters between the router and the target prompt.
  • Capability inventory: The router skill focuses on file reading and orchestration without requiring elevated system privileges or network access.
  • Sanitization: Content from the referenced repository files is adopted directly to provide the requested workflow guidance.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 06:58 AM
Security Audit — agent-trust-hub — agentic-workflows