OpenParable
Skip to content
← Back to skills

Slate Ar Stabilize

Testing & Quality
slate-ar-stabilize

Slate Ar Stabilize is an Agent Skill for testing & quality published in the udecode/plate repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Slate v2 behavior-stability mini-skill. Runs existing editor behavior gates, routes repeated failures to slate-patch, and keeps looping until the surface is stable or blocked.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/udecode/plate/blob/HEAD/.agents/skills/slate-ar-stabilize/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/slate-ar-stabilize, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Slate AR Stabilize

Handle $ARGUMENTS.

Use this when the user wants editor behavior regressions gone: selection, typing, copy/paste, IME, focus, undo/redo, navigation, browser route behavior, or native interaction parity.

This is an expert override. The normal broad workflow is slate-ar-perfect <surface>, which should stabilize behavior before perf.

Contract

Use autogoal for broad stabilization requests. For one exact bug, route directly to slate-patch.

If no behavior surface is named, infer the riskiest current behavior surface from slate-ar-status, recent failing gates, and current changed code. If that cannot be inferred, recommend one concrete gate command instead of asking the user to choose among skills.

Start with slate-ar-status, then:

  1. run the narrowest existing behavior gate through slate-ar-gate;
  2. if the same valid gate fails twice, stop gating and fix with slate-patch;
  3. if no oracle exists, create it with slate-patch or tdd;
  4. rerun the focused gate after each fix;
  5. broaden to the next relevant gate only after focused proof is green.

Completion

Stop when:

  • the named behavior gate is green repeatedly enough for the risk;
  • known regressions in scope have focused tests;
  • fixes are verified by focused commands;
  • broad behavior proof is green when the touched surface justifies it.

Do not route to perf until stabilization is green.

Handoff

Report failed signatures, fixes, gates run, repeat count, remaining skipped behavior families, and next owner.