primer

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md:10-14 instructs runtime retrieval of free-form design text from a public web endpoint via curl, which would be read into the LLM context; that downloaded content is outsider-authored.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs a runtime fetch via curl of https://api.agent-ds.oday-bakkour.com/v1/systems/primer/design.md which would retrieve a remote design.md containing the "Agent Prompt Guide" and tokens that directly control agent instructions, so this URL is flagged.