biopharma-patent-landscape
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, including web search results and patent full-text metadata (titles, abstracts, and claims) obtained via
patent.fetch. This content is interpolated into agent prompts for summarization and technical analysis, creating a surface for indirect prompt injection. - Ingestion points:
web.searchresults (SKILL.md Step 1) andpatent.fetchcontent (references/legacy-company-patent-workflow.md Steps 5, 7, 8, 9). - Boundary markers: Absent. The skill does not explicitly define delimiters or instructions to ignore embedded commands in the processed patent text.
- Capability inventory: File system write access via
python.run(Excel generation), network search capabilities viaweb.search, and patent data tools. - Sanitization: None described. The skill relies on direct keyword matching and LLM summarization without explicit escaping of external content.
- [DYNAMIC_EXECUTION]: The skill uses
python.runto execute logic for Excel formatting, data parsing (dates/IPC/legal status), and target extraction. These operations use static code templates provided in the instruction files and utilize standard libraries for structured data processing. The execution is scoped to the skill's primary purpose of patent portfolio management.
Audit Metadata