Score ML model governance risk tiers using rubric-defined axes and controls.
npx skills add https://github.com/gtylee/CodexGAS --skill risk-tiering
Please help me install this Agent Skill. Skill: Risk Tiering Source: https://github.com/gtylee/CodexGAS/tree/main/modelgas/skills/risk_tiering Command: npx skills add https://github.com/gtylee/CodexGAS --skill risk-tiering
This Skill enables governance teams to assign a consistent risk tier to ML models by scoring them across key axes (financial impact, reliance, usage, complexity, and mitigation strength) and linking the results to required controls.
Run risk_tiering on input IR and rubric data, then review the produced findings and control mapping.
High-intent search queries and answers about installing and using this skill.
ML model governance risk tiering scores models across financial impact, reliance, usage, complexity, and mitigation axes. The total score maps to a specific tier using a predefined rubric, linking to mandated controls.
Evidence-driven risk tiering evaluates ML model axis scores and total scores against a predefined rubric and control mappings. It produces structured schema-compliant findings with rationale for audits and incident reviews.
Control mapping translates the assigned risk tier into mandated controls across product, data, and deployment scenarios. It uses control mappings to provide evidence-backed justification for required safeguards.
Model risk tiering requires input IR and rubric data to execute the scoring process. It produces schema.json compliant findings containing axis_scores, total_score, tier, and rationale fields.
Yes, risk tiering applies to model governance workflows, audits, and incident reviews where evidence-backed tiering determines required safeguards across product, data, and deployment scenarios.
The best way is using an axis-based risk scoring approach that computes scores per axis and a total tier using a predefined rubric. This ensures consistent evaluation of model reliance and complexity.
SimHackerCreate weighted rubrics and score outputs against defined criteria.
pradyumnasagarGuide authors through creating and validating ORS SKILL.md files.
OWASPVerify AI applications against the AISVS framework and generate a compliance checklist.
levnikolaevichCompute Usefulness Scores and classify tests as KEEP, REVIEW, or REMOVE.
