By nushey
Spec-Driven Development orchestrator — Init, Tech Lead, Developer, Verifier subagents (full SDD) plus a Mini-SDD planner + developer subagent for smaller tasks. Harness-neutral so it works with Claude Code, Gemini CLI, Codex, and any agent that reads AGENTS.md.
Mini-SDD Developer subagent. Cold-context implementer for a Mini-SDD `plan.md`: honors Bootstrap, executes all tasks, commits, runs declared validations, reports back. Harness-neutral. Never pushes.
SDD Developer. Implements exactly ONE task file from .spec/<slug>/tasks/ or .spec/<slug>/fixes/. Adds tests alongside implementation when the task requires them. Commits the change on the current feature branch using conventional commits. Never pushes. Invoke once per task during the SDD Implement phase or for a fix task during failure recovery.
SDD Init & Preparer Agent. Refines the Orchestrator's intake.md into a polished scope.md: well-formed user stories, observable acceptance criteria, formalized Reference Files, and a complete contract for the Tech Lead. Also guarantees AGENTS.md exists at the project root. Invoke ONLY during the SDD Phase 1 (Init).
SDD Tech Lead. Defines the technical design for the whole feature AND decomposes it into atomic, value-oriented tasks the developer can execute sequentially. Produces design.md, tasks.index.md, and one file per task under tasks/. During failure recovery, produces fix tasks under fixes/ without redesigning. Invoke during the SDD Design+Tasks phase and during failure recovery. Does NOT write production code.
SDD Verifier (QA + PR gate). Runs tests, reviews all commits against scope acceptance criteria and design, checks for overengineering and convention violations, writes verify.md, and (on PASS) pushes the feature branch and opens a pull request. Never merges. Invoke during the SDD Verify phase.
Prompt for generating an AGENTS.md file for a repository
Mini-SDD Planner. Merges Init and Tech Lead roles into a single workflow. Creates a unified plan.md for smaller features or fixes. Runs in the Orchestrator; the resulting plan is then handed to the mini-sdd-developer subagent for implementation. Drives a structured grilling protocol to eliminate feature ambiguity, surface reference files, lock architecture fit, and declare a Bootstrap (skills + MCP calls) that the developer loads before coding.
Mini-SDD Flow. A leaner version of SDD for smaller features, bug fixes, or refactors. Triggered by "/mini-sdd". Planning runs in the Orchestrator (mini-sdd-planner skill); implementation is delegated to a single mini-sdd-developer subagent so the implementation context starts clean and honors the Bootstrap (skills + MCP calls) declared in the plan. Harness-neutral: works with Claude Code, Gemini CLI, Codex CLI, and any agent harness that reads AGENTS.md.
Standard for writing PR descriptions in the SDD Verifier phase. Defines structure and tone: value-oriented, concise, technical detail only when it matters to the reviewer.
Spec-Driven Development flow. Trigger when user requests SDD, spec-driven implementation, says "/sdd", "let's spec this", "use SDD", or asks to plan a feature end-to-end through init → design+tasks → implement → verify. Claude becomes the Orchestrator and delegates to sdd-init, sdd-tech-lead, sdd-developer, and sdd-verifier subagents.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Spec-Driven Development orchestrator. Turns any feature prompt into a disciplined init → design+tasks → implement → verify flow, with specialized subagents coordinated via a .spec/<feature-slug>/ folder.
Cross-agent by design: ships with both a Claude Code plugin and a Gemini CLI extension, and the skills/agents use harness-neutral wording (e.g. "load the declared skills", "invoke the MCP tool by name") so they also work with Codex CLI or any other agent harness that reads AGENTS.md.
AGENTS.md is law; ensured up-to-date by sdd-init at the start of every run.| Component | Type | Purpose |
|---|---|---|
sdd | Skill | Orchestrator skill — invoked by /sdd <feature> |
create-agentsmd | Skill | Authors AGENTS.md for fresh repos (fallback used by sdd-init) |
pr-creation | Skill | PR body standard used by the Verifier — value-oriented, minimal technical noise |
writing-skill | Skill | Standard for structured technical documentation used by all agents |
sdd-init | Subagent | Preparation phase — ensures AGENTS.md is current and defines the scope.md contract |
sdd-tech-lead | Subagent | Defines technical design AND decomposes into atomic value-oriented tasks |
sdd-developer | Subagent | Implements one task, commits, fills the per-task Implementation log, never pushes |
sdd-verifier | Subagent | Runs tests, cross-checks Implementation logs against git, opens PR on PASS |
agents-md | MCP server | Scans existing repos to bootstrap or update AGENTS.md |
Claude Code (latest).
uv — required for the bundled agents-md MCP server.
Linux / macOS:
curl -LsSf https://astral.sh/uv/install.sh | sh
Windows (PowerShell):
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Git + GitHub CLI (gh) — the Verifier opens PRs via gh pr create.
Install the extension directly from the repository:
gemini extensions install https://github.com/nushey/sdd-flow
(Or point to a local directory: gemini extensions install ./sdd-flow)
/plugin marketplace add nushey/sdd-flow
/plugin install sdd-flow@sdd-flow
Restart Claude Code. The sdd skill, four subagents, and the agents-md MCP register automatically.
/sdd Add OAuth login with Google
Best for big features, complex refactors, and hard changes requiring deep validation and context isolation. Uses subagents.
/mini-sdd Fix the typo in the header and update the styles
Best for small features, bug fixes, or refactors where the full flow is overkill. Planner runs in the orchestrator; implementation is delegated to a single mini-sdd-developer subagent so the implementation context starts clean and loads the skills/MCPs declared in the plan.
sdd-init ensures AGENTS.md exists and writes scope.md. This file centralizes business intent, acceptance criteria, and style references.sdd-tech-lead writes design.md (feature-level — no file lists), tasks.index.md, and one task file per atomic unit. Each task ships with Reference Files for strict style matching.sdd-developer executes each task: reads task + design.md + curated context, commits with conventional commits, then fills the Implementation log.sdd-verifier cross-checks logs against git, runs tests, reviews Architectural Fidelity, and opens a PR on PASS.All artifacts live under .spec/<feature-slug>/. Re-running /sdd on an existing slug resumes where it left off.
fixes/ — design.md stays as-is.MIT
npx claudepluginhub nushey/sdd-flow --plugin sdd-flowReliable automation, in-depth debugging, and performance analysis in Chrome using Chrome DevTools and Puppeteer
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Complete developer toolkit for Claude Code
Intelligent draw.io diagramming plugin with AI-powered diagram generation, multi-platform embedding (GitHub, Confluence, Azure DevOps, Notion, Teams, Harness), conditional formatting, live data binding, and MCP server integration for programmatic diagram creation and management.