spec-driven_skill

This skill coordinates a spec-driven development workflow, launching phase subagents, managing approvals, and guiding transitions to ensure structured feature

83

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill nikiforovall/claude-code-rules --skill spec-driven

  • SKILL.md5.1 KB

Overview

This skill orchestrates a spec-driven development workflow that enforces Requirements → Design → Tasks → Implementation with approval gates between phases. I coordinate and launch specialist subagents for each phase, manage iterations, and never create spec files or implement work directly. The skill is ideal when you want a structured, auditable planning process guided by explicit approvals.

How this skill works

I act as the orchestrator: for each phase I launch the appropriate subagent (requirements-agent, tech-design-agent, tasks-agent, implementation-agent), wait for its output, present results for review, and only advance when you explicitly approve. I enforce sequential execution, handle user feedback by relaunching agents with revision instructions, and stop after task approval unless you request implementation. At implementation time I still launch the implementation-agent and never write code or docs myself.

When to use it

  • When you want a repeatable, gated feature planning process (Requirements → Design → Tasks → Implementation).
  • When you say “use spec-driven” or ask to follow the spec process for a feature.
  • When you need clear, reviewable artifacts and explicit approvals before implementation.
  • When multiple stakeholders must review and sign off at each phase.
  • When you want incremental task execution with strict adherence to specs.

Best practices

  • Always provide a clear feature description before launching the requirements-agent.
  • Review each agent’s output carefully and give concrete feedback for revisions.
  • Approve explicitly at each gate — approvals are required to proceed.
  • Avoid asking the orchestrator to write docs or implement tasks; it must launch subagents instead.
  • Use small, focused tasks for implementation to keep iterations fast and verifiable.

Example use cases

  • Planning a new feature that requires cross-team technical review and acceptance criteria.
  • Converting a product idea into measurable user stories with EARS-style acceptance criteria.
  • Producing a detailed implementation plan where tasks are reviewed and approved before coding.
  • Running incremental implementations where each task is executed strictly against the spec.
  • Onboarding teams to a disciplined handoff process from product requirements through implementation.

FAQ

No. The orchestrator never writes requirements, designs, tasks, or code. It must launch the correct subagent for each artifact or implementation step.

What happens if I don’t approve a phase?

The workflow pauses. Provide feedback and I will relaunch the same subagent with your revision requests until the output is approved.

Can implementation be grouped into multiple tasks per session?

No. Implementation-agent executes one task per session and must follow the spec exactly without improvisation.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
spec-driven skill by nikiforovall/claude-code-rules | VeilStrat