bid-template-fill

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). Yes — the skill asks the agent/user to collect and place sensitive values (bank account numbers, unified social credit code, contact numbers, etc.) into JSON and use them to generate filled documents, which requires the LLM to handle and potentially emit those secret values verbatim.

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

• Third-party content exposure detected (high risk: 0.85). RUNTIME path: scripts/bid_pipeline.py → parse_bid_markdown() → read_bid_file() reads the outsider-authored bid document text (PDF/DOCX/MD) and then injects extracted free text fields (e.g., buyer/agency/project name/lots/deadline/payment/warranty/delivery) into the LLM context via the agent’s prompt variables used for template filling.