implement

Implement approved spec folders with runtime verification and one triaged code review.

Updated Sep 17, 2026
One-click install
npx skills add https://github.com/hkw2028/toy-project --skill implement-hkw2028
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: implement
Source: https://github.com/hkw2028/toy-project/tree/main/.agents/skills/implement
Command: npx skills add https://github.com/hkw2028/toy-project --skill implement-hkw2028

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning an approved specification into working, verified code often breaks down: specs go stale, interrupted sessions lose context, reviews get skipped or repeated, and unverified changes ship. This Skill treats a docs/specs/SLUG/ folder as a single handoff bundle and drives it to completion with disciplined verification and review. ## Core Features & Use Cases - Handoff-driven implementation: Loads the spec, active task files, linked decisions, and approved reference artifacts before changing any source, then implements one outcome at a time through a public test seam. - Runtime verification and reconciliation: Verifies each outcome in the running product, reconciles observed behavior against acceptance criteria, and reruns complete deterministic verification before review. - Single triaged review pass: Runs one automated whole-diff code review, repairs only confirmed defects or broken acceptance criteria, records the rest, and hands off a runnable local server when the repository exposes one. - Use Case: You have an approved spec folder with four task files and want it finished end to end. The Skill implements tasks sequentially, resumes safely after interruptions, completes verification and review, and shares a reachable local URL for your review. ## Quick Start Ask the AI to implement the docs/specs/<slug>/ folder end to end, including verification, one automated review pass, and a runnable handoff.

Frequently Asked Questions about implement

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

FAQPage Schema
How do I implement a spec folder end to end with an AI coding agent?

Point the agent at the docs/specs/<slug>/ folder and ask it to implement it. The Skill loads the spec, active tasks, and approved references, implements outcomes sequentially, verifies each in the running product, and finishes with one automated review pass.

How do I resume an interrupted implementation session safely?

Ask the agent to resume the spec folder. It reconstructs current truth from the spec folder, task statuses, code, Git state, and fresh verification rather than trusting prior conversation, and confirms ownership before absorbing ambiguous uncommitted changes.

What happens when a spec conflicts with what the repository actually supports?

The Skill stops before weakening the approved product contract. It preserves current artifacts and evidence, presents the exact product decision for the user to settle through shaping, and keeps dependent tasks blocked until resolved.

Does the Skill run code review more than once after repairs?

No. Each scope gets exactly one automated review pass, including repairs. Confirmed must-fix findings are repaired and reverified with tests and runtime checks, but the reviewer is never invoked twice in the same run.

Can it hand off a running product for manual review?

Yes, when the repository exposes a local dev or preview server. It starts an isolated server without disrupting other worktrees, verifies the changed routes and states, and shares a reachable address. CLI-only repositories get a normal verified handoff instead.

What if the automated reviewer is unavailable or user-only?

Overall completion stays pending. The Skill delivers verified implementation evidence, names the exact blocker, and reports only a user command the current harness actually confirms, unless the user explicitly waives the review.