monitor-adc-patents-weekly-ls
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [SAFE]: The skill utilizes authorized MCP servers (advanced_patent_search, patent_briefing, etc.) provided by the vendor Patsnap for data retrieval. These represent legitimate vendor-owned tools and are documented for intended functionality.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external patent publications, news sources, and drug databases via MCP tools described in SKILL.md. While the skill establishes a strict reporting structure and a two-pass screening process to evaluate relevance, it lacks explicit prompt-level boundary markers (e.g., delimiters) for external content. However, this surface is inherent to the skill's primary monitoring function and is mitigated by structured analysis requirements.\n
- Ingestion points: External patent bibliographic data, news articles, and drug asset milestones enter through the Patsnap MCP tools.\n
- Boundary markers: Absent. The skill uses structured tables and categories but does not define explicit delimiters for ingested data.\n
- Capability inventory: The skill has access to network-based search tools (via MCP) and instructions to generate Markdown or HTML reports.\n
- Sanitization: The skill documents a two-pass screening process that acts as a logical filter for content relevance.
- [NO_CODE]: The skill consists of instructions in SKILL.md without additional local scripts or executables, relying entirely on remote MCP server tools for processing.
Audit Metadata