What problem does it solve? Teams need a consistent, deterministic way to assess the risk of a story during refinement and of a pull request during review, without each reviewer inventing their own criteria. This Skill applies a single quality model to produce a five-dimension classification matrix and a risk tier, so downstream gates and automation consume one authoritative result. ## Core Features & Use Cases - Dual-context classification: Classifies from the story context during refinement and from the code diff during review, with a never-lower rule so review can only confirm or raise the refinement tier. - Model-driven rules: Applies the quality model (KB defaults plus an optional tech/risk-matrix.md adoption delta) with a resolution cascade of argument override, adoption, then KB default. - Body output and tag projection: Writes a one-line verdict plus a collapsed details matrix into the card or PR body, and emits chromatic risk/cost tags only when the adoption declares a tag projection. - Use Case: During PR review, invoke it on a pull request to recompute the matrix from the diff, floor it against the refinement tier, and raise the tier to risk:red if the diff reveals a schema migration or authentication change. ## Quick Start Ask the assistant to classify the current pull request in review context and write the resulting risk matrix into the PR body.
