ljg-analogy
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [NO_CODE]: The skill consists entirely of instructional markdown files and JSON/YAML configuration. It contains no scripts, binaries, or automation code.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied scenarios which acts as an ingestion surface. However, the skill does not possess any high-risk capabilities such as file system access, network exfiltration, or command execution, making the potential for exploitation negligible.
- Ingestion points: User-provided scenarios and target domains for analogy processing.
- Boundary markers: Not explicitly implemented in the instructions.
- Capability inventory: Restricted to text processing; no system-level or network tools are invoked.
- Sanitization: No specific data sanitization routines are described.
- [EXTERNAL_DOWNLOADS]: Includes a reference to a research paper on the Gentner 1983 theory hosted on a legitimate university domain (northwestern.edu) for academic documentation purposes.
Audit Metadata