sd0x-dev-flow

Language: English | 繁體中文 | 简体中文 | 日本語 | 한국어 | Español

The autonomous development workflow engine for Claude Code.

• Zero manual gates — edit code, auto-review, auto-fix, ship
• Dual-reviewer architecture — Codex MCP + secondary reviewer in parallel, fail-closed
• ~4% context footprint — 96% of Claude's 200k window stays for your code

65 commands | 49 skills | 14 agents | 5 hooks | 12 rules | 11 scripts

Quick Start

# Install plugin
/plugin marketplace add sd0xdev/sd0x-dev-flow
/plugin install sd0x-dev-flow@sd0xdev-marketplace

# Configure your project
/project-setup

One command auto-detects framework, package manager, database, entrypoints, and scripts. Installs 12 rules + 5 hooks.

Use --lite to only configure CLAUDE.md (skip rules/hooks).

How It Works

flowchart LR
    P["🎯 Plan"] --> B["🔨 Build"]
    B --> G["🛡️ Gate"]
    G --> S["🚀 Ship"]

    P -.- P1["/codex-brainstorm<br/>/feasibility-study<br/>/tech-spec"]
    B -.- B1["/feature-dev<br/>/bug-fix<br/>/codex-implement"]
    G -.- G1["/codex-review-fast<br/>/precommit<br/>/codex-test-review"]
    S -.- S1["/smart-commit<br/>/push-ci<br/>/create-pr<br/>/pr-review"]

Loading

The auto-loop engine enforces quality gates automatically — after any code edit, Claude triggers dual review (Codex MCP + secondary reviewer in parallel) in the same reply. Findings are deduplicated, severity-normalized, and aggregated into a single gate. Hooks enforce fail-closed semantics: if the aggregate gate is incomplete, stop-guard blocks.

Detailed: Dual-Review Sequence Diagram

sequenceDiagram
    participant D as Developer
    participant C as Claude
    participant X as Codex MCP
    participant T as Secondary Reviewer
    participant H as Hooks

    D->>C: Edit code
    H->>H: Track file change
    C->>H: emit-review-gate PENDING
    par Dual Review
        C->>X: Codex review (sandbox)
    and
        C->>T: Task(code-reviewer)
    end
    X-->>C: Findings (primary)
    T-->>C: Findings (secondary)
    C->>C: Aggregate + dedup + gate
    C->>H: emit-review-gate READY/BLOCKED

    alt Issues found
        C->>C: Fix all issues
        C->>X: --continue threadId
        X-->>C: Re-verify
    end

    C->>C: /precommit (auto)
    C-->>D: ✅ All gates passed

    Note over H: Fail-closed: incomplete gate → blocked

Loading

Feature Spotlight: Dual-Reviewer Architecture

v2.0 dispatches two independent reviewers in parallel — zero single-point-of-failure:

Reviewer	Role	Fallback
Codex MCP	Primary (sandbox, full diff)	Always available
Secondary (pr-review-toolkit)	Confidence-scored review	strict-reviewer → single mode

Findings are severity-normalized (P0-Nit), deduplicated (file + issue key, ±5 line tolerance), and source-attributed (codex | toolkit | both).

Gate: ✅ Ready or ⛔ Blocked — fail-closed (incomplete gate = blocked).

When to Use

Good Fit	Not Ideal
Solo or small-team projects with Claude Code	Teams not using Claude Code
Projects needing automated review gates	One-off scripts with no CI
Codex CLI / Cursor / Windsurf users (skills subset)	Projects requiring custom LLM providers
Repos where quality gates prevent regressions	Repos with no test infrastructure

Install

Codex CLI / Other AI Agents

# Install individual skills via Agent Skills standard
npx skills add sd0xdev/sd0x-dev-flow

# Generate AGENTS.md + install hooks (in Claude Code)
/codex-setup init

Method	Tools	Coverage
Plugin install	Claude Code	Full (65 commands, hooks, rules, auto-loop)
npx skills add	Codex CLI, Cursor, Windsurf, Aider	Skills only (49 skills)
/codex-setup init	Codex CLI	AGENTS.md kernel + git hooks

Requirements: Claude Code 2.1+ | Codex MCP (optional, for /codex-* commands)

Workflow Tracks

Workflow	Commands	Gate	Enforced By
Feature	/feature-dev → /verify → /codex-review-fast → /precommit	✅/⛔	Hook + Behavior
Bug Fix	/issue-analyze → /bug-fix → /verify → /precommit	✅/⛔	Hook + Behavior
Auto-Loop	Code edit → /codex-review-fast → /precommit	✅/⛔	Hook
Doc Review	.md edit → /codex-review-doc	✅/⛔	Hook
Planning	/codex-brainstorm → /feasibility-study → /tech-spec	—	—
Onboarding	/project-setup → /repo-intake	—	—

Visual: Workflow Flowcharts

flowchart TD
    subgraph feat ["🔨 Feature Development"]
        F1["/feature-dev"] --> F2["Code + Tests"]
        F2 --> F3["/verify"]
        F3 --> F4["/codex-review-fast"]
        F4 --> F5["/precommit"]
        F5 --> F6["/update-docs"]
    end

    subgraph fix ["🐛 Bug Fix"]
        B1["/issue-analyze"] --> B2["/bug-fix"]
        B2 --> B3["Fix + Regression test"]
        B3 --> B4["/verify"]
        B4 --> B5["/codex-review-fast"]
        B5 --> B6["/precommit"]
    end

    subgraph docs ["📝 Docs Only"]
        D1["Edit .md"] --> D2["/codex-review-doc"]
        D2 --> D3["Done"]
    end

    subgraph plan ["🎯 Planning"]
        P1["/codex-brainstorm"] --> P2["/feasibility-study"]
        P2 --> P3["/tech-spec"]
        P3 --> P4["/codex-architect"]
        P4 --> P5["Implementation ready"]
    end

    subgraph ops ["⚙️ Operations"]
        O1["/project-setup"] --> O2["/repo-intake"]
        O2 --> O3["Develop"]
        O3 --> O4["/project-audit"]
        O3 --> O7["/best-practices"]
        O3 --> O5["/risk-assess"]
        O4 --> O6["/next-step --go"]
        O5 --> O6
        O7 --> O6
    end

Loading

What's Included

Category	Count	Examples
Commands	65	/project-setup, /codex-review-fast, /verify, /smart-commit
Skills	49	project-setup, code-explore, smart-commit, contract-decode
Agents	14	strict-reviewer, verify-app, coverage-analyst
Hooks	5	pre-edit-guard, auto-format, review state tracking, stop guard, namespace hint
Rules	12	auto-loop, auto-loop-project, codex-invocation, security, testing, git-workflow, self-improvement
Scripts	11	precommit runner, verify runner, dep audit, namespace hint, skill runner, commit-msg guard, pre-push gate, utils (shared lib), emit-review-gate, worktree-claude-sync, build-codex-artifacts

Minimal Context Footprint

~4% of Claude's 200k context window — 96% remains for your code.

Component	Tokens	% of 200k
Rules (always loaded)	5.1k	2.6%
Skills (on-demand)	1.9k	1.0%
Agents	791	0.4%
Total	~8k	~4%

Skills load on-demand. Idle skills cost zero tokens.

Commands Reference

Command	Description
/project-setup	Auto-detect and configure project
/feature-dev	Feature development workflow
/bug-fix	Bug/Issue fix workflow
/codex-review-fast	Quick review (diff only)
/codex-review-doc	Document review
/precommit	lint:fix → build → test
/precommit-fast	lint:fix → test (no build)
/verify	Full verification chain
/smart-commit	Smart batch commit
/push-ci	Push + CI monitor
/create-pr	Create GitHub PR
/codex-brainstorm	Adversarial brainstorm (Nash equilibrium)
/tech-spec	Generate tech spec
/pr-review	PR self-review
/codex-security	OWASP Top 10 audit

All 65 commands

Development

Command	Description
/project-setup	Auto-detect and configure project
/repo-intake	One-time project intake scan
/install-rules	Install plugin rules to .claude/rules/
/install-hooks	Install plugin hooks to .claude/
/install-scripts	Install plugin runner scripts
/codex-setup	Initialize Codex CLI infrastructure (AGENTS.md + hooks)
/bug-fix	Bug/Issue fix workflow
/codex-implement	Codex writes code
/codex-architect	Architecture advice (third brain)
/code-explore	Fast codebase exploration
/git-investigate	Track code history
/issue-analyze	Deep issue analysis
/post-dev-test	Post-dev test completion
/feature-dev	Feature development workflow (design → implement → verify → review)
/feature-verify	System diagnosis (read-only verification with dual-perspective)
/load-pr-review	Load GitHub PR review comments into session
/pr-comment	Post friendly review comments to a GitHub PR
/code-investigate	Dual-perspective code investigation (Claude + Codex independent)
/next-step	Context-aware next step advisor
/smart-commit	Smart batch commit (group + message + commands)
/git-profile	Git identity and GPG signing profile manager
/push-ci	Push (with approval) + CI monitor
/create-pr	Create GitHub PR from branch
/git-worktree	Manage git worktrees (auto-syncs .claude/)
/merge-prep	Pre-merge analysis and preparation
/smart-rebase	Smart partial rebase for squash-merge repos

Review (Codex MCP)

Command	Description	Loop Support
/codex-review-fast	Quick review (diff only)	--continue <threadId>
/codex-review	Full review (lint + build)	--continue <threadId>
/codex-review-branch	Full branch review	-
/codex-cli-review	CLI review (full disk read)	-
/codex-review-doc	Document review	--continue <threadId>
/codex-security	OWASP Top 10 audit	--continue <threadId>
/codex-test-gen	Generate unit tests	-
/codex-test-review	Review test coverage	--continue <threadId>
/codex-explain	Explain complex code	-
/seek-verdict	P2 dismiss blind verification	-

Verification

Command	Description
/verify	lint -> typecheck -> unit -> integration -> e2e
/precommit	lint:fix -> build -> test:unit
/precommit-fast	lint:fix -> test:unit
/dep-audit	Dependency security audit
/project-audit	Project health audit (deterministic scoring)
/best-practices	Industry best practices audit with adversarial debate
/risk-assess	Uncommitted code risk assessment

Planning

Command	Description
/codex-brainstorm	Adversarial brainstorm (Nash equilibrium)
/feasibility-study	Feasibility analysis
/tech-spec	Generate tech spec
/review-spec	Review tech spec
/deep-analyze	Deep analysis + roadmap
/project-brief	PM/CTO executive summary

Documentation & Tooling

Command	Description
/update-docs	Sync docs with code
/check-coverage	Test coverage analysis
/create-request	Create/update request docs
/doc-refactor	Simplify documents
/simplify	Code simplification
/de-ai-flavor	Remove AI-generated artifacts from documents
/safe-remove	Safely remove plugin assets
/skill-creator	Create new skills (external plugin)
/pr-review	PR self-review
/pr-summary	PR status summary (grouped by ticket)
/contract-decode	EVM contract error/calldata decoder
/skill-health-check	Validate skill quality and routing
/statusline-config	Customize statusline segments and themes
/claude-health	Claude Code config health check
/op-session	Initialize 1Password CLI session (avoids repeated biometric prompts)
/obsidian-cli	Obsidian vault integration via official CLI
/zh-tw	Rewrite in Traditional Chinese

Rules

Rule	Description
auto-loop	Fix -> re-review -> fix -> ... -> Pass (auto cycle)
auto-loop-project	Project-specific auto-loop overrides (user-owned, not plugin-managed)
codex-invocation	Codex must independently research, never feed conclusions
fix-all-issues	Zero tolerance: fix every issue found
self-improvement	Corrected → record lesson → prevent recurrence
framework	Framework-specific conventions (customizable)
testing	Unit/Integration/E2E isolation
security	OWASP Top 10 checklist
git-workflow	Branch naming, commit conventions
docs-writing	Tables > paragraphs, Mermaid > text
docs-numbering	Document prefix convention (0-feasibility, 2-spec)
logging	Structured JSON, no secrets

Customization: Edit auto-loop-project.md to override auto-loop behavior per project. Plugin updates won't conflict — see Rule Override Pattern.

Hooks

Hook	Trigger	Purpose
namespace-hint	SessionStart	Inject plugin command namespace guidance into Claude context
post-edit-format	After Edit/Write	Auto prettier + invalidate review state on edit
post-tool-review-state	After Bash / MCP tools	Track review state (sentinel routing, supports namespaced commands)
pre-edit-guard	Before Edit/Write	Prevent editing .env/.git
stop-guard	Before stop	Block or warn on incomplete reviews + stale-state git check (strict after install, warn in plugin runtime)

Hooks are safe by default. Use environment variables to customize:

Variable	Default	Description
STOP_GUARD_MODE	strict (installed) / warn (plugin runtime)	strict blocks stop on missing review steps; warn only warns
HOOK_NO_FORMAT	(unset)	Set 1 to disable auto-formatting
HOOK_BYPASS	(unset)	Set 1 to skip all stop-guard checks
HOOK_DEBUG	(unset)	Set 1 to output debug info
GUARD_EXTRA_PATTERNS	(unset)	Regex patterns for extra protected paths (e.g. src/locales/.*\.json$)

Dependencies: Hooks require jq. Auto-format requires prettier. Missing dependencies are handled gracefully.

Customization

Run /project-setup to auto-detect and configure all placeholders, or manually edit .claude/CLAUDE.md:

Placeholder	Description	Example
{PROJECT_NAME}	Your project name	my-app
{FRAMEWORK}	Your framework	MidwayJS 3.x, NestJS, Express
{CONFIG_FILE}	Main config file	src/configuration.ts
{BOOTSTRAP_FILE}	Bootstrap entry	bootstrap.js, main.ts
{DATABASE}	Database	MongoDB, PostgreSQL
{TEST_COMMAND}	Test command	yarn test:unit
{LINT_FIX_COMMAND}	Lint auto-fix	yarn lint:fix
{BUILD_COMMAND}	Build command	yarn build
{TYPECHECK_COMMAND}	Type checking	yarn typecheck

Showcase: StatusLine Config

Customize Claude Code's statusline — segments, themes, and colors. Install as a standalone skill:

npx skills add sd0xdev/sd0x-dev-flow --skill statusline-config

Feature	Details
Segments	Directory, Git branch, Model, Context %, Cost, >200k alert
Themes	ansi-default, catppuccin-mocha, dracula, nord, none
Engine	POSIX shell + JSON stdin + semantic color tokens
Accessibility	WCAG AA contrast, NO_COLOR support

Full documentation

Architecture

Command (entry) → Skill (capability) → Agent (environment)

• Commands: User-triggered via /...
• Skills: Knowledge bases loaded on demand
• Agents: Isolated subagents with specific tools
• Hooks: Automated guardrails (format, review state, stop guard)
• Rules: Always-on conventions (auto-loaded)

For advanced architecture details (agentic control stack, control loop theory, sandbox rules), see docs/architecture.md.

Contributing

PRs welcome. Please:

1. Follow existing naming conventions (kebab-case)
2. Include When to Use / When NOT to Use in skills
3. Add disable-model-invocation: true for dangerous operations
4. Test with Claude Code before submitting

License

MIT

Star History