setback-predictor
Pass
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates as a purely analytical tool with no evidence of malicious patterns. It defines a multi-phase process (discovery, risk modeling, early warning indicators, etc.) for auditing existing codebases and data structures related to clinical rehabilitation. The use of
$ARGUMENTSis restricted to scoping the analysis (e.g., 'readmission prediction model accuracy') and does not flow into shell execution or other dangerous sinks. All instructions emphasize non-modification of code and the protection of patient data. The 'SELF-EVOLUTION TELEMETRY' section is a standard pattern for logging execution metadata to a local hidden directory (~/.claude/projects/) and does not involve external exfiltration.
Audit Metadata