OpenParable
Skip to content
← Back to skills

Slate Ar Perfect

Development
slate-ar-perfect

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

Slate v2 broad improvement mini-skill. Uses Autogoal plus Slate AR routing to improve a named surface across quality, bugs, behavior proof, and perf without requiring the user to pick sub-skills.

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-perfect/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/slate-ar-perfect, 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 Perfect

Handle $ARGUMENTS.

Use this when the user wants a Slate v2 surface made genuinely good, not one isolated packet. This is the high-level loop for "perfect pagination", "perfect huge document perf", "perfect editor behavior", and similar asks.

This is the main "absolute best architecture / DX / API / testing / perf" workflow. The user should not need to manually call slate-ar-fast or slate-ar-stabilize after slate-ar-perfect; those are internal or expert override lanes.

Contract

Use autogoal for durable work. The goal must name:

  • the surface;
  • measurable or auditable completion threshold;
  • architecture/API/DX acceptance rows when relevant;
  • behavior and perf gates;
  • boundaries;
  • blocked stop condition.

Keep the goal short. Put the real checklist in the goal plan.

Routing

Start with slate-ar-status, then choose owners in this order:

  1. slate-ar-quality for current-state gaps across API, DX, architecture, examples, tests, and perf surfaces;
  2. slate-plan only when the remaining issue needs public API/runtime design;
  3. slate-patch for known bugs or missing behavior oracles;
  4. slate-ar-gate for existing editor behavior proof;
  5. slate-ar-perf for benchmark-backed perf targets;
  6. slate-ar-gate again for final broad no-regression proof.

Do not accept a perf win while native behavior regresses. Faster broken editor behavior is not progress.

Perf waits for behavior stability unless the perf benchmark itself is the only way to reproduce the behavior failure. If perf work exposes a correctness bug, route to slate-patch, fix it, then resume the perf target.

Completion

Stop when the surface has:

  • accepted architecture/API/DX gaps implemented, routed to slate-plan, or consciously deferred with evidence;
  • no known P0/P1 behavior regressions in scope;
  • relevant behavior gates green or explicitly N/A with evidence;
  • relevant perf targets promoted or plateaued with correctness green;
  • final broad no-regression proof green when the surface touches editor behavior;
  • goal plan completion check green.

Completion does not mean creating autoresearch-review/* branches. Stay on the source branch, normally v2, and use slate-ar-ship readiness previews unless the user explicitly asks for review branches.

Handoff

Report:

  • goal plan path;
  • surface and completion threshold;
  • packets/fixes/gates run;
  • accepted deferrals;
  • residual risks.