- Home
- Skills
- Prompt & Skill Engineering
- AI Protocol Kit: approval-first workflow for coding agents
More screenshots
Works with the AI Agents you already use
AI Protocol Kit: approval-first workflow for coding agents
It is a set of instruction files and small Python scripts that you install into a project once.
$3.99
AI Protocol Kit: approval-first workflow for coding agents
Example session with this skill installed
Set up the AI Protocol Kit in a new Python project called "Deneme". My language is English. Use the python stack and add the stack module.
- Read your context and instructions
- Compiled the ai protocol kit
Summary
The kit was installed into the new project and the language is set to English. The agent now has to translate the instruction files; a task file lists exactly which ones.
Setup output
=== AI Protokol Kiti Kurulumu ===
Proje: Deneme | Stack: python | Dil: en
Dil: en - agent'in cevirmesi gereken 16 dosya Agent_Files/CEVIRI_GOREVI.md icinde.
=== Kurulum tamamlandi ===
SIRADAKI ADIM (agent): Agent_Files/CEVIRI_GOREVI.md dosyasini uygula, sonra ceviri_dogrula.py calistir.
What was created
| Item | Purpose |
|---|---|
CLAUDE.md, AGENTS.md, .cursorrules, .cursor/rules/ | Rules the agent follows every session |
Agent_Files/ | Tracking files, session and finalize scripts |
proje_yapilandirma.json | Project settings; ceviri_durumu is bekliyor until translation is verified |
Agent_Files/CEVIRI_GOREVI.md | The translation task for the agent, with the files and the items that must not change |
Next steps
- The agent translates the listed files into English.
- It runs
python Agent_Files/ceviri_dogrula.py; on success the task file is removed. - It runs the session-start command once to confirm everything works.
Connects securely to your tools. The creator never sees your data.
About this skill
What it does
AI Protocol Kit gives a coding agent the same working agreement in every session. It is a set of instruction files and small Python scripts that you install into a project once. After that the agent:
- Starts each session cheaply. It loads a short context card and the current module's tracking file instead of re-reading the whole backlog or chat history.
- Asks before it codes. One focused round of questions, a short summary of the plan, then it waits for your explicit approval.
- Edits surgically. Small edits are free, medium edits need a one-line summary, large rewrites need approval plus a backup commit first.
- Tracks work honestly. Every change is logged as coded, awaiting your test, or confirmed. Nothing is reported as "tested" unless it was actually tried.
- Finishes with one command.
finalizeupdates the tracking files and makes a local backup commit. Push, force operations and branch deletion always need your approval.
Who it is for
- You start a new project with an AI editor and want rules from day one.
- Your agent rewrites large files, skips approval, or loses track of what was tested.
- You want a token-saving workflow instead of a growing pile of chat context.
What you get
- Instruction files for Claude Code and Cursor (
CLAUDE.md,AGENTS.md,.cursorrules,.cursor/rules/) - Tracking files: backlog, completed work, known issues, pending tests, per-module notes
agent_oturum_yonetici.pywithbaslangic,finalize,arsivandtamcommands- Optional stack modules: Flutter, Unity, Godot, Web, Python, plus templates for data security, store/legal checks and splitting rules into small files
- A setup script that asks for project name, stack and language, fills the placeholders and installs everything
- A lessons document explaining why each rule exists, taken from real project incidents
Language
The templates, file names and script output are in Turkish. During setup you choose your language; the agent then translates the instruction files for you and answers in your language. A verification script rejects untranslated files and any damaged command or placeholder. File names and script output stay Turkish.
Requirements
Python 3 and git. Windows, Linux and macOS command examples are included.
Not included
It does not guarantee error-free agent behavior; it reduces mistakes through discipline. The rules are strict by design (approval before code). Relax them in the project's CLAUDE.md if you want a faster, less supervised flow.
Requires Python 3 and Git. Automation script commands and file names are in Turkish. Not suitable for one-off scripts or tiny projects where overhead exceeds utility.
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
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 15 days ago
Requires
Permissions
Writes instruction and tracking files into the project folder you choose and runs the bundled Python scripts. finalize makes a local git commit; it never pushes. No network access and no environment variables are used.