write-code

Implements new features end to end through a risk-adaptive TDD lifecycle with review and publication gates.

4|1|Updated Jul 31, 2025
One-click install
npx skills add https://github.com/alvis/.agents --skill write-code-alvis
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: write-code
Source: https://github.com/alvis/.agents/tree/main/plugins/coding/skills/write-code
Command: npx skills add https://github.com/alvis/.agents --skill write-code-alvis

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coordinating a full feature implementation—design, skeleton, tests, fixes, refactoring, review, lint, and publication—across many steps and state files is error-prone, and this Skill owns that entire TDD lifecycle with explicit gates and state reconciliation. ## Core Features & Use Cases - Risk-adaptive topology: Classifies work into semantic risk tiers, keeping Tier 0/1 with one owner, adding independent review for Tier 2, and Tech Lead coordination for Tier 3. - Composed phase pipeline: Orchestrates setup, draft, completion, fix, refactor, documentation, review, lint, and final test/type/build gates in a defined order. - State and history discipline: Reconciles task status in state.md, seals scoped-save manifests, and delegates commits, finalization, and PR publication to dedicated skills. - Use Case: Ask it to implement a new API endpoint with acceptance criteria; it drafts the skeleton, completes implementation and tests, runs review and lint, and prepares a scoped commit and PR. ## Quick Start Use the write-code skill to implement the new user settings endpoint with validation, tests, and documentation according to the approved ticket.

Frequently Asked Questions about write-code

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

FAQPage Schema
How do I implement a new feature with tests using an AI coding workflow?

Provide an instruction with scope, behavior, and acceptance criteria, and the skill runs a TDD lifecycle: draft skeleton, complete implementation and tests, fix failures, refactor, then review, lint, and final test and build gates before publication.

When should I use write-code instead of fix or refactor skills?

Use write-code for new functions, features, modules, or endpoints implemented end to end. Route diagnosed failures to fix, green structural cleanup to refactor, skeleton-only work to draft-code, and accepted production stubs to complete-code.

Can write-code resume an interrupted implementation?

Yes, pass --resume with a resolved work ID whose state.md defines unfinished scope. It reads state/working.md and the task table, then schedules only runnable leaf tasks; a missing local root must go through the takeover skill.

Does write-code commit and create pull requests directly?

No, it never invokes commit, push, or PR commands directly. History shaping is delegated to the commit skill with a sealed scoped-save manifest, per-commit QA to finalize-commits, and publication to the pr skill.

What happens if requirements are vague or lack acceptance criteria?

The skill rejects vague requirements without acceptance criteria. Runtime-producing projects also require a configured test framework, while declaration-only type utilities may use typecheck or consumer build routes instead.