The Agent Skills Directory

[SAFE]: The skill is authored by Patsnap and correctly utilizes authorized vendor-owned MCP tools (e.g., patsnap-patent-search, patsnap-zhihuiya-analysis-report) and official domains (eureka.zhihuiya.com) for its intelligence gathering functions.
[SAFE]: The workflow incorporates robust quality control gates that explicitly prohibit the use of hallucinated or estimated data, mandating verification against primary sources such as SEC filings and official corporate announcements.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates an ingestion surface for indirect prompt injection as it processes external web content and user-provided documents to generate reports.
Ingestion points: Processes reference_doc files and web.search/web_fetch results (SKILL.md).
Boundary markers: No explicit delimiters are defined for external content.
Capability inventory: Performs web reading, patent database queries, and generates a local HTML report.
Sanitization: Implements a multi-step verification process (Step 11) designed to validate factual accuracy against known sources before report finalization.

pharma-company-intelligence