What problem does it solve? Verifying that a Persona and its Skills behave correctly after a prompt, model, or runtime change is hard to do repeatably, since results can be contaminated by shared context or previous conversations. This Skill runs each Persona–Skill pair in a fresh, isolated low-reasoning Codex task, evaluates the response against a defined contract, and archives the temporary task afterward. ## Core Features & Use Cases - Isolated Test Matrix: Generates one test case per live Persona–Skill pair via node eval/isolated-persona-skill.mjs matrix, excluding prototype records. - Contract-Based Evaluation: Checks Persona–Skill identity, reasoning level, isolation cleanliness, response-contract completeness, quality signals, and unauthorized mutation claims with evaluate <result.json>. - Task Lifecycle Management: Creates fresh Codex tasks, waits on bounded batches of up to eight, captures final responses, and archives tasks after capture. - Use Case: After changing a model or prompt, run a full regression pass across every live Persona–Skill pair, then review the sanitized result bundles under eval/results/isolated/ for conformance verdicts. ## Quick Start Use $persona-skill-conformance to test all live Persona–Skill pairs in fresh low-reasoning tasks and archive the temporary tasks.
