implement

Implements work from a spec or tickets with tests, quality gates, and code review.

Updated Jul 29, 2026
One-click install
npx skills add https://github.com/MaiconGambini/opencode-harness-guide --skill implement-maicongambini
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: implement
Source: https://github.com/MaiconGambini/opencode-harness-guide/tree/main/skills/implement
Command: npx skills add https://github.com/MaiconGambini/opencode-harness-guide --skill implement-maicongambini

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning a spec or set of tickets into verified, committed code requires coordinating TDD, typechecking, test runs, quality gates, and review — this Skill orchestrates that entire implementation workflow so nothing is skipped. ## Core Features & Use Cases - Spec-driven implementation: Builds the work described in a spec or tickets, applying TDD at pre-agreed seams. - Layered verification: Runs typechecking and single test files regularly, the full test suite once at the end, then the quality gate in local mode and a code review at the router-assigned depth. - Scheduler lane mode: When dispatched as a bounded specialist by a scheduler, it edits and verifies without committing, runs the gate with a lane label, and returns a metric table so the scheduler can reconcile parallel lanes. - Use Case: A scheduler dispatches three parallel lanes to implement separate tickets; each lane edits code, runs a lane-labelled local quality gate, and returns metrics, while the scheduler performs the single reconciled commit. ## Quick Start Implement the work described in the current spec or tickets, run the local quality gate, review the changes, and commit to the current branch.

Frequently Asked Questions about implement

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

FAQPage Schema
How do I implement work from a spec or tickets with automated verification?

Describe the work from the spec or tickets and let the Skill implement it using TDD at pre-agreed seams. It runs typechecking and single test files regularly, the full test suite once at the end, then a local-mode quality gate and code review before committing.

When should the quality gate run in local mode versus full mode?

Local mode is used by individual implementation runs and parallel lanes because mutation and e2e testing in full mode cost minutes per lane. The scheduler runs full mode once over the reconciled tree after all lanes finish.

Why does a scheduler-dispatched lane not commit its changes?

The scheduler is the sole committer and reconciles all parallel lanes into one local commit. If a lane committed directly, parallel lanes would race on git and break the single-committer invariant.

Can parallel lanes write to the quality-thresholds baseline file?

No. Lanes must never write agent-os/quality-thresholds.json because concurrent writes to one JSON file lose updates. Only the scheduler's full-mode run updates baselines, preserving the one-writer-per-file invariant.

What must a lane return to the scheduler after finishing?

A lane must run the quality gate with --mode local --label <lane-id> and include the gate's metric table in its result. Returning only a claim like tests pass gives the scheduler nothing concrete to reconcile.