Execute

SkillProductivity

Execute is a skill that lets an AI agent carry an approved task through to working, verified code. Once a plan or change has been approved, the agent implements it and checks that the result actually works before reporting back.

Use Execute in Claude, ChatGPT or Ahel Desktop

Free. Sign in, add Execute and connect your AI. About a minute.

Also: Claude Code · Cursor · Codex

Then ask your AI: use the Execute skill

Details

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Have an agent that supports loading skills.

ExecuteStart free

What your AI can do with it

  • Carry an approved task through to working code
  • Verify that the produced code works before finishing
  • Turn an approved plan or change into an implementation
  • Report back only after the result is confirmed working

Getting started

  1. Have an agent that supports loading skills.
  2. Add the execute skill to the agent's available skills.
  3. Approve a task or plan with the agent, then ask it to execute it.

What this skill tells your AI

The instructions your AI receives, as published by yeachan-heo/oh-my-claudecode in skills/execute/SKILL.md and read by ahel’s review.

Use this skill when the work is understood and the job is to build it.

This is the canonical execution workflow. autopilot, ralph, ultragoal, ultrapilot, pipeline, and swarm route here.

Goal

Take a task from agreed intent to working code, with evidence that it works.

Workflow

  1. Confirm the task is clear enough to build. If it is not, plan first.
  2. Break the work into independent units; run genuinely independent units in parallel.
  3. Implement the smallest correct change per unit, reusing existing utilities and patterns.
  4. Verify as you go, not only at the end.
  5. Report what changed, what was verified, and what remains.

Scale

Match the machinery to the task:

  • Single unit — implement directly, verify, done.
  • Several independent units — delegate to executor agents in parallel.
  • Long-running or unbounded — keep a durable task list and continue until the list is empty.
  • Needs coordinated parallel workers — use team.

Do not spin up coordination for work that one focused pass would finish.

Rules

  • Prefer deletion over addition when behavior is preserved.
  • Do not add dependencies without an explicit request.
  • Keep diffs small and reversible.
  • Placeholder TODOs, test.skip, and stub tests are blockers, not progress.
  • Authoring and approval are separate passes — do not self-approve; hand off to review or verify.

Completion

Before claiming done:

  • No pending tasks
  • Tests pass, or failures are reported plainly
  • Verification evidence collected

Output

  • Files changed
  • What was implemented
  • Evidence it works
  • What is still open

Signals

GitHub stars
39k
Forks
4k
Last commit
Oct 2026
Hacker News mentions
20

Questions

What does the execute skill do?
It lets an agent carry an approved task through to working, verified code, checking that the result works before reporting back.
When should it be used?
After a task or plan has been approved and the next step is to implement it and confirm the code works.
Advanced
Item type
skill
Key
execute-yeachan-heo
Source
github.com/yeachan-heo/oh-my-claudecode