agentic-workflows

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly loads prompt/skill files at runtime from the external repository "github/gh-aw" (e.g., via @.github/aw/instructions.md), and those fetched files are then followed directly to control the agent's behavior, so this is a runtime remote-content dependency that controls prompts.