code

Implement one planned step test-first and commit it as a single reviewable diff.

Updated Feb 12, 2026
One-click install
npx skills add https://github.com/kieranpotts/skills --skill code-kieranpotts
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: code
Source: https://github.com/kieranpotts/skills/tree/main/skills/code
Command: npx skills add https://github.com/kieranpotts/skills --skill code-kieranpotts

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning a planned, already-designed step into working code often leads to scope creep, untested changes, and bloated diffs that are painful to review. This Skill enforces disciplined execution: one step, one test-first implementation, one clean commit. ## Core Features & Use Cases - Test-first implementation: Works in red-green-refactor cycles, writing one failing test and one minimal implementation per behavior, with unit and integration tests alongside the code. - Scope locking: Quotes the step verbatim, measures the finished diff against it, and stops rather than guessing when work is ambiguous, too large, or still needs designing. - Convention discovery: Reads the project's documented coding, testing, and commit conventions (or infers them from surrounding code and git history) instead of assuming paths or formats. - Use Case: After a planning skill decomposes a feature into numbered steps, invoke this Skill with "implement step 3 of the plan" to get a single tested, reviewable commit for exactly that step. ## Quick Start Ask the agent to implement step 3 of the delivery plan, test-first, and commit it as one reviewable diff.

Frequently Asked Questions about code

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

FAQPage Schema
How do I implement a planned step test-first with an AI agent?▼

Invoke the skill with a prompt like "implement step 3 of the plan" or "build this change, test-first". The agent quotes the step's scope, sets up a fast test loop, works in red-green-refactor cycles, and commits one reviewable diff.

What inputs does a test-driven coding workflow need before starting?▼

It needs one scoped unit of work whose design is already settled, such as a numbered step of a delivery plan or a small standalone change. Work that still needs designing or decomposing is rejected rather than guessed at.

Can this skill handle work that has not been broken down yet?▼

No. If the step is ambiguous, too large for one reviewable diff, or still needs designing, the agent stops and reports it for splitting. Use a planning or decomposition skill upstream first.

Does the agent push commits or open pull requests after coding?▼

No. Work stops at the local commit. Pushing, opening a review, releasing, and starting the next step are left to the caller, keeping each invocation to exactly one reviewable diff.

What happens when the code being changed has no existing tests?▼

The agent adds characterization tests that pin down current behavior before changing it, committed separately from the change itself. This ensures modifications are verified against a trustworthy baseline.

When is it acceptable to skip test-first discipline?▼

For trivial changes like config tweaks, renames, or copy edits, test-first may be skipped. Explicit spikes or prototypes also skip it, but the prototype must be thrown away and re-implemented properly afterward.