agentflow

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). SKILL.md describes a workflow where workers read/write task state via comments on an external Kanban/PM tool (outsider-authored free text from those comment threads), which would be ingested into the agent’s LLM context at runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The installation steps instruct to "git clone https://github.com/UrRhb/agentflow.git" and copy its prompts/skills into ~/.claude, meaning remote repository content is fetched and then loaded as agent prompts/skills that directly control the agent at runtime (https://github.com/UrRhb/agentflow.git).