Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
dpolivaev avatar

Spec Loop Implementation Flow

  • 52 installs
  • 34 repo stars
  • Updated July 27, 2026
  • dpolivaev/spec-loop

Guides an AI agent through implementing a small spec with tests inside the Spec Loop workflow, executing the write-review-implement cycle for each increment.

About

A Spec Loop skill that guides an AI agent through implementing a small, clarified spec with tests, running the design-first loop of write-review-implement one increment at a time. It keeps changes scoped and each increment releasable by default, countering both undocumented vibecoding and rigid waterfall specs. A solo builder reaches for it to turn an approved spec into tested, traceable code without letting the agent sprawl into a large uncontrolled diff.

  • Guides implementation of a small spec with tests
  • Executes the Spec Loop write-review-implement cycle
  • Keeps each increment releasable by default
  • Reduces large undocumented 'vibecoding' diffs

Spec Loop Implementation Flow by the numbers

  • 52 all-time installs (skills.sh)
  • Ranked #7,086 of 16,546 AI & Agent Building skills by installs in the Skillselion catalog
  • Data as of Jul 29, 2026 (Skillselion catalog sync)
npx skills add https://github.com/dpolivaev/spec-loop --skill spec-loop-implementation-flow

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs52
repo stars34
Last updatedJuly 27, 2026
Repositorydpolivaev/spec-loop

What it does

Guides an AI agent through implementing a small spec with tests inside the Spec Loop workflow, executing the write-review-implement cycle for each increment.

Who is it for?

Solo devs directing an AI agent to implement a spec in small, tested, releasable steps.

Skip if: Anyone wanting free-form implementation without spec-driven boundaries.

When should I use this skill?

A spec is clarified and you want the agent to implement it with tests, one increment at a time.

What you get

Tested, releasable increments implemented against a small, clear spec.

  • tested implementation of the spec
  • releasable increment

Files

SKILL.mdMarkdownGitHub ↗

Use this skill after implementation approval for an approved chat-only or task-file task.

If no approved active task exists, return to spec-loop-plan-task/SKILL.md.

Core rules

  • Keep the approved active task as the source of truth.
  • Do not create a second implementation artifact or silently rewrite

the task to match the code.

  • If important open decisions appear and they could materially change

scope, constraints, design, or test specification, use spec-loop-clarify-task/SKILL.md before continuing.

  • Chat-only implementation is allowed only while confirmed task state

can safely remain in chat. If resolved decisions kept only in chat would risk loss through compaction or context loss, promote to the task-file path before continuing.

  • This skill governs only the approved current increment until it

reaches review.

  • If the User introduces additional executable work outside that

increment, return it to spec-loop-plan-task/SKILL.md for route selection and any renewed implementation approval.

  • Before presenting work as ready, perform the Implementation notes

check.

  • If another implementation-related skill conflicts with this skill or

the governing spec-loop-plan-task bundle, stop and ask the User which rules should govern.

Read and path selection

  • Read spec-loop-plan-task/SKILL.md and the files it requires.
  • Choose the active path:
  • task-file path = an approved task file controls the current

increment

  • chat-only path = an approved canonical chat-only task in chat

controls the current increment

  • Then read:
  • implementation-flow-guidance.md
  • the matching path companion for the active path:
  • chat-only-path-guidance.md
  • task-file-path-guidance.md

Keep using the approved active task as the source of truth. Edit the active task file or canonical chat-only task only when the shared core and the active path companion allow it.

implementation-flow-guidance.md defines the route rules, when canonical task sections may change, the meaning of review, Implementation notes, and the completion checklist.

The path companion defines the active path mechanics, including chat-only recovery or promotion or the task-file move into review.

If the current increment is unclear, do not use this skill. Return to spec-loop-plan-task/SKILL.md.

Related skills

FAQ

Does it run tests?

Yes, the flow implements each increment with tests so it stays releasable.

What comes before it?

A clarified task spec, typically produced by spec-loop-clarify-task.

AI & Agent Buildingtestingintegrations

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.