coding-lifecycle

name

coding-lifecycle

description

Use when a coding-capable agent is starting or finishing implementation work. At start, clarify intent, scope, context, and verification strategy. At finish, run appropriate quality checks, assess impact, capture lessons, and produce a concise completion report. Trigger for both implementation intake and implementation closeout. Load with read_file on .kilocode/skills/coding-lifecycle/SKILL.md (ignore the absolute path in the location tag).

Coding Lifecycle

This skill is the portable coding-work wrapper. Use the entrypoint that matches the current phase.

Entrypoint: Start coding work

1. Read the task or delegation request.

   • Identify the objective, target behavior, known files or components, constraints, and success criteria.
   • If the task came from a coordinator, preserve the coordinator's scope and acceptance criteria.

2. Discover relevant context.

   • Identify dependencies, side effects, risky areas, and affected routes, models, services, utilities, or external APIs.

3. Choose a verification strategy.

   • Select the lightest reliable verification for the change: focused tests, lint or type checks, browser/manual checks, CLI checks, inspection-only review, or a combination.
   • If new tests are appropriate, first look for similar tests to emulate or extend.

4. Begin implementation.

   • Follow coding-guidelines throughout implementation.
   • Return to the calling workflow's implementation step with scope, context, risks, and verification strategy established.

Entrypoint: Finish coding work

1. Review the changed scope.

   • Confirm the implementation matches the requested behavior and stayed within scope.
   • Perform impact analysis for dependent routes, models, services, utilities, tests, docs, and configured workflows.

2. Run quality checks.

   • Run the project-appropriate quality checks that match the verification strategy.
   • Resolve relevant failures introduced by the change before proceeding.
   • If a check cannot be run, record why and use the next-best available verification.

3. Capture reusable lessons when warranted.

   • If the task revealed a non-obvious fix or repeatable pattern, use the available learning mechanism to save it.

4. Complete the task.

   • Report changed files/components, rationale, verification performed, and risks or follow-ups.