mini-sdd

Orchestrate a spec-driven workflow from task prompt to approved plan and implementation.

1|Updated Apr 18, 2026

Search Tags:#planning #implementation #verification #spec-driven development #GitHub pull request #sequential commits

Ready to run

npx skills add https://github.com/nushey/sdd-flow --skill mini-sdd

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: mini-sdd
Source: https://github.com/nushey/sdd-flow/tree/main/skills/mini-sdd
Command: npx skills add https://github.com/nushey/sdd-flow --skill mini-sdd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Mini-SDD reduces the risk of ad-hoc or drift-prone development on smaller tasks by enforcing a disciplined plan and verification loop before implementation.

Core Features & Use Cases

Lean spec-driven workflow: Converts a user prompt into a focused plan and sequential implementation steps for small features, bug fixes, or refactors.
Interactive planning with guardrails: Uses an orchestrator-led grilling protocol to produce a single plan file as the source of truth.
Single subagent implementation: Delegates implementation to exactly one subagent to ensure the plan’s declared skills and MCP calls are explicitly honored.
Verification and reporting: Runs final checks (including tests and acceptance criteria) and returns a structured report without pushing or merging.

Quick Start

Run mini-sdd to generate an approved plan and then implement and verify the change according to that plan.

Frequently Asked Questions about mini-sdd

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

What is spec-driven development and how does it prevent implementation drift?▼

Spec-driven development enforces a structured plan and verification loop before writing code. By creating an approved plan file as the source of truth, it prevents ad-hoc implementation drift on smaller features, bug fixes, and contained refactors.

How do I plan and implement small code changes sequentially?▼

You can plan and implement small changes by running an orchestrator-led interactive planning phase to generate a .spec/<slug>/plan.md file, then delegating execution to a single subagent that makes sequential commits based on that approved plan.

Can I use this spec-driven workflow for larger feature refactors?▼

This workflow is designed for smaller features, bug fixes, and contained refactors. Larger tasks may exceed its scope, which relies on a single subagent and a focused planning phase to manage implementation.

Does the implementation phase handle GitHub pull request creation and merging?▼

No, the workflow performs verification and returns a structured report without pushing or merging. It focuses on local implementation, running final checks against acceptance criteria, and generating sequential commits.

How does the orchestrator ensure declared skills and MCP tool calls are used during implementation?▼

The orchestrator delegates implementation to exactly one subagent, which is explicitly required to honor the declared skills and MCP tool invocations specified in the generated plan during the execution phase.

What's the best way to verify acceptance criteria after a small bug fix?▼

Use a spec-driven workflow that runs final checks, including tests, against the acceptance criteria defined in your plan file. It then returns a structured report confirming verification before any code is pushed.