The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of patent claims, descriptions, and legal statuses, which can contain embedded instructions designed to influence the AI's behavior during report generation.
Ingestion points: Patent metadata and full-text evidence retrieved via the advanced_patent_search and patent_briefing MCP connectors (documented in SKILL.md).
Boundary markers: The skill implements a confidentiality_boundary and requires explicit user confirmation of research scope in Stage 1.
Capability inventory: Generates persistent HTML and DOCX report files and performs network requests via MCP servers.
Sanitization: The instructions mandate escaping all retrieved content, allowing only verified safe links, and explicitly prohibit loading remote scripts or executing retrieved content in the final deliverables.
[EXTERNAL_DOWNLOADS]: The skill refers to official Patsnap MCP connectors for search and mining operations.
Source: Official vendor marketplace links are provided for open.patsnap.com domains.

conduct-patent-research-ip