What problem does it solve? Evaluating plugins with a shared generic rubric produces meaningless scores, and deciding pass/fail thresholds after seeing results invites bias. This Skill guides you through deriving what 'good' means for a specific plugin from its own profile, writing it into a gated rulers.md document, and getting a person to approve it before any scoring begins. ## Core Features & Use Cases - Custom rubric derivation: Builds per-plugin rulers with qualitative dimensions (every level named, low end first) and quantitative dimensions (tool-computed facts with human-set thresholds and stated reasons). - Registry validation: Uses ruler_record and ruler_affirm to reject dimensions with missing levels, empty thresholds, or figures that do not exist on the plugin's facts sheet. - Approval gating: Writes rulers.md via document_write, presents it with document_present, and records human agreement with document_approve before judging can start. - Use Case: After profiling a delivery-method plugin, you define a 'document depth' qualitative dimension with five named levels and a 'tool fit' quantitative threshold over never_called counts, then get stakeholder sign-off before the judge scores any artifact. ## Quick Start Run the zz-plugin-define stage to derive evaluation rulers for this plugin from its profile and write them into rulers.md for approval.
