academic-pipeline

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly retrieves and interprets public third‑party content (e.g., via DOI/API/manual_pdf retrieval of entries from literature_corpus[] and web searches) as part of its integrity and claim-alignment workflows—see the SKILL.md and agents/claim_ref_alignment_audit_agent.md (Step 2 “Reference retrieval”) and integrity_verification_agent examples—so untrusted web content can materially influence judge decisions and pipeline gating.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches external reference content at runtime (e.g., DOI link https://doi.org/10.1186/s41239-026-00585-x) and injects the retrieved excerpt into the claim-audit/judge prompts that determine audit outcomes, so this external URL content is used at runtime to directly control agent prompts and is a required dependency.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)