ljg-mechanism
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a purely instructional and reasoning-based framework designed for phenomena analysis. It does not include any executable code, scripts, or system modification commands.
- [DATA_EXPOSURE]: The skill provides explicit guidelines in Workflows/DiscoverMechanism.md to protect user privacy, instructing the agent to never send original private materials to external search services during the verification process.
- [PROMPT_INJECTION]: Instructions are focused on methodology and output structure. There are no attempts to bypass safety filters, extract system prompts, or override agent behavior through injections.
- [INDIRECT_PROMPT_INJECTION]: Although the skill is designed to process external internet data, it lacks high-risk capabilities (such as file-writing or code execution) and emphasizes source verification and competing explanation analysis, which serves as a defensive layer against malicious data ingestion.
Audit Metadata