Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsAI Coding Workflow
A repeatable workflow for AI coding agents in any repository: explore, plan, approval before changes, test-first build, verification, review and a release checklist.
Free
AI Coding Workflow
Example session with this skill installed
I need a standard coding workflow for this React project. We use vitest for tests and want an approval step before any files are modified. Can you set this up?
- Read your context and instructions
- Compiled the ai coding workflow
Example of the first turn. The skill reads the repository first, then asks one question at a time and waits for each answer.
I looked at the repository and found a vitest setup (npm run test) and a CONTRIBUTING.md. I'll ask four short questions, one at a time.
Question 1 of 4: What repeats? I see many component updates and bug fixes. Should the routine focus on features and bug fixes first?
(You answer. The skill then asks where you want to be asked, what "done" means, and whether npm run test is the command that proves a change works.)
What it proposes before it writes anything
| File | Action | Purpose |
|---|---|---|
WORKFLOW.md | Create | The routine: Explore, Plan, Build, Verify, Review, Hand over. |
AGENTS.md | Create | A short pointer, read by Codex and Copilot's coding agent. |
CLAUDE.md | Update | The same pointer, read by Claude Code. |
.cursor/rules/workflow.mdc | Create | The same pointer, always applied in Cursor. |
It writes nothing until you say yes.
Next steps
- Answer the four questions.
- Approve the files above.
- Try a dry run on a tiny task, such as a typo in a comment, and watch the agent follow
WORKFLOW.md.
Connects securely to your tools. The creator never sees your data.
About this skill
AI coding agents are fast, but they often change code before planning, skip tests, or push work nobody has checked. This skill sets up a short, repeatable workflow in your repository that your agent reads at the start of every task. It asks the agent to plan first, report each step and stop to ask where a wrong move is costly.
What it does
- Sets guardrails: says exactly when the agent must stop and ask before changing, pushing or publishing code.
- Asks for test-first builds: write a failing test, then the smallest change that makes it pass.
- Asks for real verification: run the change like a user would and check a neighbouring feature, instead of relying only on passing tests.
- Tidies handovers: small commits in your repository's own message style, and a pull request only when you say so.
- Adds a release checklist: version, release notes, build, tag, publish, verify and a rollback line.
How it works
- Interview: the skill asks four short questions: what work repeats, where you want to be asked, what "done" means, and which command proves a change works.
- Preview: it shows you the routine and the files it would change, and writes nothing until you say yes.
- Files: it creates one
WORKFLOW.mdand adds a short pointer inAGENTS.md,CLAUDE.mdand an always-applied Cursor rule. - Routine: your agent follows Explore, Plan, Build, Verify, Review and Hand over on each task.
- Dry run: it offers to try the routine on a tiny task and fixes any rule that felt slow or wrong.
Works with
Any repository and any agent that reads AGENTS.md, CLAUDE.md or Cursor rules, including Claude Code, Cursor, GitHub Copilot's coding agent and Codex. An optional reference covers the Agent World workflow file format.
Why not just prompt it yourself?
A prompt you type fades as a conversation grows. These rules live in files in your repository that your agent reads at the start of each task, and your team shares and edits them like any other file.
Use cases
- Asking agents to stop before touching production-critical files.
- Prompting a failing test before every bug fix.
- Keeping one routine across a team that uses different AI coding tools.
- Putting multi-step releases (version, tag, publish, rollback) in one checklist.
Known limitations
- An agent can still ignore the routine. This skill does not enforce it, and it has not been measured to improve code quality. What it gives you is more control and predictability.
- You need to answer four questions about your repository.
- The agent needs permission to create and edit markdown files in the repository root and in
.cursor/rules.
Free and open source (Apache License 2.0): https://github.com/hodkep/agent-world/tree/main/docs/skills/ai-coding-workflow
An agent can still ignore the routine: the skill sets it up but cannot enforce it. It has not been measured to improve code quality, so expect more control and predictability, not better code. It needs a few answers from you about your repository, including the command that proves a change works. The agent needs permission to create and edit markdown files in the repository root and in .cursor/rules. It does not run tests, builds or CI itself.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
2 uses
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 7 days ago
- Free with an Agensi account
Permissions
File Scopes
Reads the repository to find its test and build commands, then writes WORKFLOW.md and short pointer lines in AGENTS.md, CLAUDE.md and a Cursor rule file. It shows you the files first and writes nothing until you say yes. It runs no commands, uses no network and touches no other files.