wiki-query
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed with a robust verification workflow for information retrieval. It mandates citations for every claim and explicitly forbids the modification of raw source files, ensuring data integrity during the query process. The creation of new content through 'synthesis' is strictly opt-in, preventing unauthorized or silent modifications to the wiki index.
- [PROMPT_INJECTION]: The skill processes untrusted data from wiki pages and raw sources, presenting a surface for indirect prompt injection.
- Ingestion points: Data is ingested from 'wiki/index.md', candidate wiki pages via 'wiki_read_page', and raw source files via 'wiki_read_source'.
- Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore or isolate potential instructions embedded within the wiki or source content.
- Capability inventory: The toolset includes 'wiki_search', 'wiki_read_page', 'wiki_list_sources', 'wiki_read_source', and 'wiki_write_page'.
- Sanitization: No specific sanitization or filtering of the ingested content is described; however, the requirement for inline citations and the verification checklist provide a layer of oversight.
Audit Metadata