Stop That Shit
A hook and skill guard for AI coding agents in Codex, Claude Code, OpenCode, Hermes Agent and Pi that blocks four unrequested behaviors: scope creep, unneeded defensive code, exceeding a stated intent such as review-only, and repeating work that already has an answer.
Stop That Shit is a hook and skill guard that stops an AI coding agent from padding a task with unrequested work. It defines four behaviors to intercept: scope creep, where fixing one problem turns into refactoring an unrelated module; unnecessary defensive code, such as a checksum nobody reads or a compatibility layer for an imagined future; violating a stated intent, such as editing files during a read-only review; and task thrashing, redoing work that already answered the current question with no new information.
Instead of open-ended rules, it works through an ordered check: what the task actually needs to deliver, whether an existing capability already covers it, what specific gap remains, whether a piece of defensive code still serves a purpose, and whether the result has already been verified. Necessary work, such as a data migration with its callers and compatibility tests, must still be completed in full; the guard removes only what is not needed.
It is validated against 18 published bad and good judgment cases covering read-only review, necessary callers, migrations for already-published data, and file, dependency and subagent boundaries, with recorded test runs against Codex and GPT models. It installs as a plugin for Claude Code, Codex, OpenCode, Hermes Agent CLI and Pi, or as a skill-only install with no enforcement.