duolabmeng6/aicodecoreworkflow — explained in plain English
Analysis updated 2026-05-18
Add a structured research, plan, approve, execute, review workflow to Claude Code or Codex.
Require an explicit approval step before an AI coding agent starts changing files.
Give a coding agent a consistent process for handling feature requests and bug fixes.
| duolabmeng6/aicodecoreworkflow | abdullahselek/viperc | chaelsoo/hollow | |
|---|---|---|---|
| Stars | 80 | 80 | 80 |
| Language | — | Objective-C | C |
| Last pushed | — | 2024-05-14 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | easy | easy | moderate |
| Complexity | 1/5 | 2/5 | 3/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Just copy or symlink the skills/workflow folder into your agent's personal skills directory.
AICodeCoreWorkflow is a shared skill file that adds a structured workflow to AI coding assistants like Claude Code and Codex. Instead of jumping straight into writing code, the skill has the assistant follow a sequence of steps: research the problem, come up with an idea, write a plan, wait for the user to approve that plan, carry out the work, then review the result. The plan is the one required approval point in the whole process. Once approved, the plan locks in the goal, the general approach, what is in and out of scope, how the change should behave from the outside, and the main risks. It does not lock down every implementation detail. Within each step the assistant is free to choose how it investigates the codebase, how deeply it reasons, which tools it uses, and how it actually writes the code. To install it, a user sends a short message to Claude Code, Codex, or any other agent that supports Agent Skills, asking it to install the skill directly from the GitHub repository link. The assistant is expected to fetch the repo, find the skills/workflow folder inside it, copy that folder into its own personal skills directory, keep automatic invocation turned on, confirm the skill was installed correctly, and tell the user how to call it directly if needed. Once installed, the assistant automatically uses this workflow for tasks that involve making real code changes. It can also be called explicitly, for example by typing /workflow followed by a description of the feature in Claude Code, or $workflow in Codex. Plain explanations, simple questions, and read only code review do not trigger the workflow automatically unless the user asks for it directly. The repository itself is small: a skills/workflow folder containing a SKILL.md file and an agents/openai.yaml configuration file. Claude Code decides when to use the skill based on the description written in SKILL.md, while Codex uses a setting in the YAML file to allow automatic use. If automatic installation does not work, the README also explains how to manually copy or link the skills/workflow folder into the right directory for each assistant.
A shared Agent Skill that makes Claude Code or Codex research, plan, get approval, then implement before touching real code.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.