wiki-claims-resweep

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly dispatches extracted inline URLs to the cogni-claims claim-verifier which performs WebFetch + LLM comparison of the live source (see SKILL.md Steps 4 and links to cogni-claims/agents/claim-verifier.md), meaning arbitrary public URLs cited in wiki pages are fetched and interpreted as part of the workflow.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill dispatches extracted source URLs (the inline http(s)://... links found in wiki pages) to the cogni-claims claim-verifier which, at runtime, fetches each source URL and injects the fetched page content into the LLM comparison/verification prompts, so arbitrary external http(s) URLs control the model input.