llm-wiki-conflict-resolver
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted external data (wiki pages and raw sources) which creates a surface for indirect prompt injection. 1. Ingestion points: Content is ingested from 'Conflicting pages', 'Supporting source pages', and 'raw sources' (SKILL.md). 2. Boundary markers: The skill lacks explicit technical boundary markers or 'ignore' instructions for embedded data, relying on logical human-in-the-loop procedures. 3. Capability inventory: The agent is instructed to read conflicting pages and sources, with the capability to write patches and log entries to the filesystem (e.g., 'wiki/log.md'). 4. Sanitization: There is no mention of sanitizing, escaping, or validating external content before it is interpolated into the agent's context or applied as a patch.
Audit Metadata