What problem does it solve? Agents often reuse previously read files, banked test results, or worker reports without knowing whether intervening writes, environment changes, or elapsed time have invalidated them. This Skill replaces guesswork with a categorical decision table that outputs reuse, spot-check, re-verify, or accept-labeled-risk for any piece of prior evidence. ## Core Features & Use Cases - Volatility Classification: Assigns each artifact to a class (stable, write-coupled, env-coupled, clock-coupled, or report) based on observable properties, not confidence feelings. - Signal-Based Decision Table: Checks intervening writes, named invalidators, decision leverage, and discriminator availability, then applies a first-match-wins table with a restrictive default for unclassified artifacts. - Inline Output Record: Emits a compact freshness record (artifact, class, signals, decision) directly inside the report or decision that consumed the evidence. - Use Case: Before relying on a test result from two hours ago to approve a release, run the contract to determine whether the result is still valid, needs a cheap spot-check, or must be fully re-verified. ## Quick Start Ask the agent to run /freshness-contract on the evidence you are about to reuse, for example by saying "is this test result from earlier still current enough to base the release decision on?"
