dialysis-prism
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE]: The skill consists entirely of markdown documentation and test cases in JSON format. It does not include any scripts, executables, or system commands.
- [SAFE]: The content describes a mental model for problem-solving (Dialysis Prism) based on the book "Cognitive Dividend". It does not attempt to access sensitive files, perform network operations, or influence the underlying agent's security boundaries.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user descriptions of problems. While this represents an ingestion surface for untrusted data, the skill possesses no capabilities (such as file writes, network requests, or code execution) that could be exploited via injection. The risk is negligible.
- Ingestion points: User prompts describing repetitive problems (processed in SKILL.md steps).
- Boundary markers: None present.
- Capability inventory: No code or system-level capabilities detected.
- Sanitization: Not applicable as the skill performs no programmatic operations.
Audit Metadata