necessity-gatelisted

<SUBAGENT-STOP> If this agent was dispatched to perform only a specific subtask, skip this skill. </SUBAGENT-STOP> <EXTREMELY-IMPORTANT> Do not bypass this gate by expert judgment. If the work "obviously" seems needed, still run the gate. This friction protects novice-day and bad-day agent behavior. Starting a new task without this gate is a rule violation. </EXTREMELY-IMPORTANT> # necessity-gate necessity-gate checks whether a new task, file, audit cycle, skill change, or follow-up is justified by real evidence and recovery value. It blocks padding, speculative cleanup, manufactured follow-ups, and scope creep. ## Contents - [Triggers](#triggers) - [Escape Hatch](#escape-hatch) - [Procedure](#procedure) - [1. State The Work](#1-state-the-work) - [2. Classify The Problem](#2-classify-the-problem) - [3. Assess Harm If Skipped](#3-assess-harm-if-skipped) - [4. Assess Regression Risk](#4-assess-regression-risk) - [5. Decide](#5-decide) - [Output Format](#output-format) - [Habitual Failure Patterns](#habitual-failure-patterns) - [Relationship To Verification](#relationship-to-verification) - [What This Gate Does Not Block](#what-this-gate-does-not-block) - [References](#references) ## Triggers Run this gate when you are about to: - define a new task, skill, or file - start a new audit or verification cycle - edit frontmatter, body, or `calls` for a production skill - add a "follow-up", "open issues", or "improvements" section to a report - register new tasks wi