tiangong-kb-patent-search

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly accepts api_key / patent_api_key and states explicit JSON fields (including those keys) are passed as CLI flags, which can require the LLM to insert secret values verbatim into generated commands.

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/patent_search.sh, the user-supplied JSON_INPUT fields query/input/claim (lines 91-93 and 135-140) are inserted into the CLI research search --query <query> path (lines 130-140), which causes the downstream Tiangong patent search system to ingest that free text at runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes "npx @tiangong-ai/cli@latest" at runtime (SKILL.md and scripts/patent_search.sh), which will fetch and execute remote npm package code when run, so it is an external runtime dependency that executes remote code.