adc-landscape-monitoring
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 information retrieved from external web searches and pharmaceutical databases to generate a landscape monitoring report, creating a surface for indirect prompt injection through the ingestion of untrusted data.
- Ingestion points: Data is ingested via multiple search tools including
ls_clinical_trial_result_search,ls_patent_search, andmcp_web-search__web_searchas specified in SKILL.md. - Boundary markers: The instructions provide guidelines for objective reporting, such as mandating the use of raw values and prohibiting subjective modifiers to reduce the risk of obedience to embedded instructions.
- Capability inventory: The skill generates local HTML files (
adc_brief_*.html) and utilizes network-based search tools. - Sanitization: While the skill enforces link validation and structured CSS tagging, it lacks explicit input filtering or escaping for the content retrieved from external search results.
- [DYNAMIC_EXECUTION]: The skill generates an HTML report that includes embedded JavaScript logic (
filterCards,filterByTag) for interactive filtering. This involves the runtime generation of executable script content within the generated output file based on retrieved data.
Audit Metadata