Skip to content
Agentmart
Submit a tool

execute

Implement an approved spec or clearly defined piece of work.

Skill by KodingDev

IndexedMIT★ 5
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/kodingdev-execute

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Implement an approved spec or clearly defined piece of work.

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/execute

What's inside

Skill name
execute
Description
Implement an approved spec or clearly defined piece of work.
License
MIT
Instructions
60 lines in SKILL.md · view on GitHub

Files (3)

  • SKILL.md
  • openai.yaml · agents/
  • implementer-prompt.md

SKILL.md (first lines)

# Execute

Implement the scope the user explicitly authorized. Invocation is approval to begin ordinary in-scope implementation; do not ask the user to approve entering the workflow again.

## Invariants

- Do not claim completion without observing risk-appropriate verification.
- Do not trust delegated completion reports without inspecting the diff and evidence yourself.
- Preserve unrelated user changes and keep the diff inside the approved scope.
- Do not delete a migration source until every caller uses and verifies the replacement. If full cutover is not achievable, leave the legacy path intact and report the partial state.
- Do not push, deploy, delete data, force-push, or perform another unrequested external or irreversible action.

## Process

### 1. Ground the work

Read the approved spec or supplied requirements, project instructions, relevant code, and current worktree state. Resolve ordinary reversible implementation details yourself. Ask only when a missing decision would materially change architecture, scope, risk, or an irreversible action.

### 2. Plan proportionately

Break multi-part work into dependency-ordered tasks. Keep a small change inline. Use `meridian:delegate` only for genuinely independent or context-heavy work where isolation materially improves execution; agent count is not evidence of diligence.

### 3. Implement and verify

For each task:

1. Search for existing interfaces, utilities, tests, and conventions before adding competing logic.
2. Implement the smallest durable change that satisfies the requirement.
3. Exercise the highest practical real seam: focused tests, typecheck, build, CLI, service, or browser behavior as appropriate.
4. If behavior fails unexpectedly, use `meridian:debug` rather than stacking speculative patches.
5. Check the diff against scope and the spec's User Constraints.
6. If a spec exists, append a concise Progress Log entry containing the task outcome, verification, deviations, and open concerns that a future agent could not recover from the diff.

### 4. Finish

Run the relevant full project checks after focused verification. Formal `meridian:review` and the commit flow are user-owned; do not run them automatically. When a commit message is written, `meridian:commit` holds the conventions.

Report:

- what changed;
- verification commands and observed results;
- any acceptance criterion that remains unverified;
- any known material concern; and
- relevant unrelated failures without treating them as introduced defects.

## Completion

- **Implementation complete:** every requested acceptance criterion is delivered and verified to the task's risk level, and no relevant required check fails.
- **Implementation incomplete:** an acceptance criterion remains undelivered or unverified.
- **Implementation blocked:** a failure or missing human-only decision prevents delivery or meaningful verification.

An unrelated pre-existing failure is reported accurately but does not change the headline unless it prevents verification of the requested work.

## Progress Log

When a spec exists, append one pickup-ready entry after each verified task and one completion entry at the end. Record outcomes, decisions not obvious from the diff, deviations, and open concerns. Do not turn the log into a tool-call journal. Do not stage or commit the spec.

Set the driving spec or sketch's **Status** to `IN PROGRESS` when implementation starts. `SHIPPED <sha>` belongs to whoever commits, since the sha does not exist yet here.

From KodingDev/meridian, MIT.

More from KodingDev/meridian

Signals

Stars
5
Forks
0
Open issues
0
Last commit (this folder)
Jul 25, 2026 (3 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from KodingDev

Cite this page

execute skill. Agent Mart. https://agentmart.dev/skills/kodingdev-execute (last updated 2026-10-24)