research-debug

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

• Third-party content exposure detected (high risk: 0.90). The skill's workflow explicitly performs WebSearch of public sites (see SKILL.md Phase 1 Step 1.1 "Web Research" and README examples listing arXiv/OpenReview/Substack links) and injects those web_search_findings into downstream agent prompts (e.g., the Task prompt's "Context from web research: {web_search_findings}"), so untrusted third‑party content is fetched and used to drive decisions and fixes.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly runs WebSearch and then injects the resulting web_search_findings into Task(...) prompts at runtime, so externally fetched content (e.g., https://arxiv.org/pdf/2503.14476) can directly control agent instructions and is a required dependency of the workflow.