AI Code Core Workflow skill
A skill that runs a task from research to review and stops only for plan approval, with the other stages going without pauses.
- 108 ★ on the repository
- Last commit
Minimal access
Why this level
- The skill changes the agent working order and itself runs no commands or network calls.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/ai-code-core-workflowDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill changes the agent working order and itself runs no commands or network calls.
This is third-party code. Review the repository files before installing.
What it does
AI Code Core Workflow is a single `skills/workflow` skill for Claude Code and Codex. It steps in on tasks that change code and splits the work into six stages: research, ideation, plan, waiting for approval, execution and review. Approval fixes the goal, approach, scope, external behavior and main hazards but does not freeze implementation details. Within stages the agent chooses its own research method, depth of reasoning and tools. Invocation is explicit: `/workflow` in Claude Code and `$workflow` in Codex. It also triggers automatically from the `SKILL.md` description, and for Codex through `allow_implicit_invocation: true` in `agents/openai.yaml`.
Who it is for. Developers working in Chinese who want a disciplined agent loop that stops at the plan.
Good fit when
- A task touches several files and a plan is needed before the first edit.
- The agent rushes into coding and review has to be requested separately.
- A team wants one working order for both Claude Code and Codex.
- A large feature is led through stages with fixed scope.
Not a fit when
- The question is an explanation or a short answer: the skill stays off by default.
- The code only needs reading without changes: read-only review is also outside automatic mode.
- The task is small and a plan with approval would slow it down.
Example request
/workflow implement email sign-in: research the project and give a plan first, edit only after my approval.Limitations
The README and interface are in Chinese. Review quality depends on the model and the project. Plan approval remains the only mandatory step, so the agent passes the other stages without stops and a flawed idea is only caught in the plan. Outside Claude Code and Codex the skill is not documented. For installation the repository link is handed to the agent, and manual copying is needed only if the agent fails: place `skills/workflow` in `~/.claude/skills/` or `~/.agents/skills/`.
How to disable. Delete the `workflow` folder from `~/.claude/skills/`, `~/.agents/skills/` or `$CODEX_HOME/skills/`. If installed by symlink, deleting the link is enough.
FAQ
How do I invoke it manually?
In Claude Code with `/workflow` plus a task description, in Codex with `$workflow`.
When does it start on its own?
On tasks that change code. Simple questions and read-only code review do not start it unless called explicitly.
What does approval lock in?
Goal, approach, scope, external behavior and main hazards. Implementation details stay free.
Where do I install manually?
To `~/.claude/skills/workflow` for Claude Code and `~/.agents/skills/workflow` or `$CODEX_HOME/skills/workflow` for Codex.
What does a symlink install give?
Updating the repository is enough: the skill picks up changes. A copy has to be transferred again.
Discussion
No questions or reviews yet.
Related
SkillsSkill and plugin for Codex and Claude Code: the agent drives a project through a plan-to-maintain loop with human gates.
A Claude Code skill that delegates tasks to OpenAI's Codex CLI for code analysis, review and edits
Eleven skills that take an idea through requirements, spec and plan to a verified pull request with an independent review
Cole Medin's skills for the prime, plan, implement, validate loop: planning, implementation, review, commit and PR in one chain.
lencx skills
lencx-skills
Two skills for Claude Code and Codex: keel works through architecture questions, coding-protocol keeps code edits inside the task
A workflow template for Claude Code and five more agents: a backlog task, a diff review and a commit.
README in short
The repository holds one skill with `SKILL.md` and a Codex config. The README sets out installing by a phrase to the agent, a manual fallback and auto-trigger rules.