CodeStable

English · 中文

An AI coding workflow for serious software engineering

Tired of OpenSpec's flimsiness, Oh-My-OpenAgent's over-engineering, and Superpowers' fragmentation — I built a lightweight, human-in-the-loop AI harness from scratch.

---

Install

Codex plugin marketplace:

codex plugin marketplace add codestable/CodeStable
codex plugin add codestable@codestable

Claude plugin marketplace:

/plugin marketplace add codestable/CodeStable
/plugin install codestable@codestable

skills CLI:

npx skills@latest add codestable/CodeStable/plugins/codestable

If your skills CLI does not discover the plugin entity through the marketplace catalog, use the deep-scan fallback:

npx skills@latest add codestable/CodeStable/plugins/codestable --full-depth

The CodeStable plugin only packages cs / cs-* skills under plugins/codestable/skills/; the repository root no longer keeps standalone skill directories.

Upgrade

After a new release, check CHANGELOG.md for the version changes, then refresh through the entry point you used to install.

Codex plugin marketplace:

codex plugin marketplace upgrade codestable
codex plugin add codestable@codestable

The current Codex CLI has no separate plugin update subcommand; marketplace upgrade refreshes the Git marketplace snapshot, and plugin add installs the current version from that refreshed snapshot.

Claude plugin marketplace:

/plugin marketplace update
/plugin update codestable@codestable

Restart Claude Code after updating so the new plugin version is applied.

skills CLI:

npx skills@latest add codestable/CodeStable/plugins/codestable --skill '*' -g

The bare skills CLI update currently uses different discovery rules for plugin manifests and generic directories, so it can misclassify existing sibling skills as deleted. Upgrade with the full-package reinstall command above instead. It synchronizes every cs* skill from plugins/codestable; for a project-scoped install, omit -g and run it in that project. Do not replace only the root cs skill: runtime refresh also requires the matching cs-onboard skill and its tools. After the global plugin upgrade, explicitly run /cs-onboard --mode refresh-runtime in every onboarded project to refresh and verify its repo-local runtime immediately. If you skip the command, the next CodeStable preflight compares .codestable/runtime-manifest.json with the current plugin version and refreshes automatically when the manifest is missing, the version or runtime capabilities do not match, and managed paths are clean; it does not scan repositories in the background. managed-paths-dirty, a repository that is not onboarded, or an incomplete skeleton stops the refresh instead of forcing an overwrite.

One command to start working:

/cs-onboard

For daily use, when you don't know which skill fits, call the root entry:

/cs

cs classifies whether you want execution, advice, or an overview. Action requests dispatch to the target skill in the current run; advice requests only recommend. Ambiguous requests get one focused question.

---

Why

I was building a new harness agent (MA) — vibe-coding at first, just writing designs and requirements while AI wrote the code. It carried most features, until Codex repeatedly failed on a problem I thought was simple, making the same mistake in the same place. That's when I knew the project needed a workflow to keep moving.

I surveyed OpenSpec, SuperPowers, Oh-My-OpenAgent — none felt right:

• OpenSpec — too thin, no compounding, specs too abstract for humans to read
• SuperPowers — no process discipline, you never know which one to use
• Oh-My-OpenAgent — too heavy, philosophically treats "human intervention = failure"

CodeStable's goal is to solve real software implementation and coding problems for serious engineering — not to coin a new term or chase trends.

---

The core difference: what gets orchestrated

Mainstream AI coding frameworks — Superpowers, CCW, Oh-My-OpenAgent — are all doing the same thing:

Orchestrating agents better. Get them to team up, collaborate, brainstorm, run pipelines, hand off automatically. The entity at the center is always the Agent.

CodeStable goes the other way:

What gets orchestrated isn't agents — it's the lifecycle of the software itself. The entities at the center are the elements that make up software: every requirement, every architectural decision, every feature, every bug, every constraint left in history.

	Agent-orchestration camp	CodeStable
Core entity	Agent / Role / Team	Requirement / Architecture / Feature / Issue / Decision
Main question	How do agents divide work, hand off, coordinate?	How do requirements, constraints, decisions get recorded, retrieved, reused?
Where state lives	Agent sessions / message buses / queues	The .codestable/ file tree in your project (readable by both humans and AI)
Pain it solves	One agent isn't enough; need coordination to scale	Software complexity overflows context; tacit knowledge gets lost; requirements drift
Role of humans	The less the better — full automation is the ideal	Human-in-the-loop — the programmer owns the whole; AI is an efficient executor

Neither direction is wrong.

If your task is "run an end-to-end automated pipeline with AI" or "have multiple agents debate a plan," the agent-orchestration camp fits better.

If your task is "maintain serious software that iterates over years" or "make sure a requirement written today can still be accurately recalled three months later" — then CodeStable's software-element-centric model fits better.

I built CodeStable because I believe the chaos of software engineering isn't really about agents not being strong enough — it's about elements not being organized. No matter how strong the agent, it can't save a project that's lost its requirements, architecture, and history.

---

Design: entities + flows

CodeStable models real coding work as a set of entities and flows.

Entities

Entity	Slug	What it does
Requirement	requirements	User stories + domain glossary (CONTEXT.md) + architecture decisions (ADRs). The escape hatch when code rots
Epic	epic	Large demand entry such as "I want a permission system"; user-facing docs call it epic while v1 still stores internal artifacts under .codestable/roadmap/
Goal	goals	Bounded start/end: write a start report, then let AI iterate autonomously with subagent functional acceptance before completion
Feature	feature	Engineering execution where human and AI share responsibility for design, implementation, QA, and acceptance
Issue	issue	Bug records and fixes after something should already work
Refactor	refactor	Behavior-preserving cleanup when code rots (beta)
Compound	compound	The compounding-engineering knowledge base: pitfalls, tricks, decisions, and investigation notes

Flows

Flow	Recommended main entry	Notes
Feature delivery	cs-feat	Risk-based lanes: Quick implements/tests/reviews directly; Standard stays in the current run with design/implementation/review/inline acceptance; Goal alone creates a long-range goal package with QA/acceptance
Epic delivery	cs-epic	Plan a large demand, review it, design child features, prepare a goal package, then dispatch a visible goal driver (print /goal as fallback)
Goal achievement	cs-goal	Bounded start/end → interview/grill + start report → autonomous implement/validate/iterate → subagent functional acceptance
Issue fixing	cs-issue	End-to-end report → analyze → fix → cs-code-review
Refactoring	cs-refactor	Behavior-preserving cleanup; standard or fastforward mode, followed by cs-code-review
External docs	cs-docs	Developer guides, user guides, and API references; hygiene stays in cs-docs-neat

cs-code-review is the cross-cutting quality gate at the tail of execution flows, before commit. At a phase or milestone boundary, use cs-docs-neat to reconcile .codestable/, README/docs, CLAUDE.md / AGENTS.md, and agent memory so docs do not drift from code.

---

Skill catalog

Recommended main entries

Group	Skill	Purpose
Root	cs	Action requests dispatch to the target skill in the current run; advice requests only recommend.
Onboard	cs-onboard	Install CodeStable into a repository
Requirements & domain	cs-req / cs-domain	Capture capability intent, domain terms, ADRs, and context topology
Epic	cs-epic	Large demand planning, review, child feature design, and goal package
Brainstorm	cs-brainstorm	Triage fuzzy ideas into feature, epic, or brainstorm notes
Goal	cs-goal	Autonomous iteration from a bounded start state to acceptance
Feature	cs-feat	Quick / Standard / Goal feature workflow; Goal is opt-in for long-range execution
Issue	cs-issue	End-to-end issue workflow
Refactor	cs-refactor	Behavior-preserving refactor workflow
Review	cs-code-review	Cross-cutting read-only implementation review gate
Audit	cs-audit	Scan for bugs, security, performance, maintainability, and architecture drift
Feedback	cs-feedback	Explicitly capture current-session incidents/triage; upload only after preview confirmation
Knowledge	cs-keep / cs-note	Capture durable knowledge or short startup-critical notes
External docs	cs-docs	Developer guides, user guides, and API references
Docs hygiene	cs-docs-neat	Sync .codestable/, README/docs, agent entries, and memory

Long-term compatibility entries

Old skill names remain usable but only enter the corresponding main workflow:

• Feature: cs-feat-design / cs-feat-design-review / cs-feat-impl / cs-feat-qa / cs-feat-accept / cs-feat-ff
• Issue: cs-issue-report / cs-issue-analyze / cs-issue-fix
• Refactor: cs-refactor-ff
• Docs: cs-doc-tutorial / cs-doc-api
• Epic: cs-roadmap / cs-roadmap-review / cs-roadmap-impl-goal

See SKILL_CATALOG.en.md for the full catalog. In daily use, call /cs when you are unsure.

---

Workflow at a glance

CodeStable is layered and event-driven:

cs
└── cs-onboard
    ├── cs-req / cs-domain
    ├── cs-epic          # user-facing epic; internally still roadmap storage
    ├── cs-goal
    ├── cs-brainstorm
    ├── cs-feat     -> cs-code-review
    ├── cs-issue    -> cs-code-review
    ├── cs-refactor -> cs-code-review
    ├── cs-docs
    ├── cs-feedback
    └── cs-keep / cs-note / cs-docs-neat

How to read it:

• cs classifies the intake mode before the target. Action requests dispatch to the target skill in the current run; advice requests only recommend. It never routes users to deprecated stage skills.
• cs-feat, cs-issue, and cs-refactor resume from repository facts. cs-feat first selects Quick, Standard, or Goal from task risk: Quick stays minimal, Standard completes in the current run with inline acceptance, and only Goal uses a long-range driver plus standalone QA. cs-issue and cs-refactor stop at their review, blocking, or user-confirmation checkpoints.
• cs-epic prepares planning and goal packages, then dispatches a visible goal driver; v1 still writes .codestable/roadmap/.
• cs-code-review is the cross-cutting gate; cs-docs-neat handles hygiene; cs-docs writes outward docs.
• cs-feedback explicitly captures a local-private current-session evidence package; public issue upload remains separately confirmed.
• Old stage skills are long-term compatibility entries for historical users.

See WORKFLOW.en.md for the compact diagram.

---

Runtime structure

After /cs-onboard, project artifacts aggregate under .codestable/. Python tool scripts run from the installed cs-onboard skill package instead of being copied into each repo.

.codestable/
├── attention.md
├── requirements/  roadmap/  goals/
├── features/  issues/  refactors/
├── audits/  brainstorms/  feedback/  compound/
├── gates/
└── reference/

requirements/ is the long-lived archive, roadmap/ is planning, dated work-item directories bundle one workflow, and compound/ is the single knowledge sink. A skill is an independent install unit: cross-skill references must be released by cs-onboard into project-local .codestable/reference/, never read from a sibling skill package. See WORKFLOW.en.md for the directory contract.

---

Design philosophy

CodeStable takes the opposite philosophy from OMO:

• OMO says: any human intervention is a failure signal
• CodeStable says: the programmer is in the loop of software coding — you may not understand the black-box implementation, but you must own the whole, and dive in when needed

Software architecture must be evolvable, observable, controllable.

This may matter less as AI gets stronger, but right now this makes programmers comfortable in reality — and that's the value.

CodeStable is modeled for real-world development scenarios, aiming to handle common dev problems through a closed-loop system. Most existing frameworks model around AI, not around humans. I think their authors have strong AI-driving skills but aren't seriously building software — they lack the basic ability to organize requirements and design, and they lack respect for code implementation.

---

Roadmap

CodeStable adapts to model capability. If a future model nails a module reliably, that module gets removed.

• Refactor flow needs hardening (cs-refactor is still beta)
• …

Issues welcome — share your real-world dev pain and refactoring experience.

---

MIT License · by @liuzhengdong