brainstorm

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). These URLs include direct raw script downloads and a personal GitHub repo that the README instructs you to curl/save a Python script from—downloading and running raw code from an unverified individual repository is a moderate–high malware risk.

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

• Third-party content exposure detected (high risk: 0.78). SKILL.md requires Google Search grounding and merges the resulting web-grounded “verified facts” (outsider-authored public web content) into the LLM prompts/context via Claude reading Flash’s saved response (Phase 0.5, “Merge: Claude reads Flash's response + own WebSearch results”).