Run a transparent, repeatable engineering workflow with requirement analysis, implementation planning, fine-grained task tracking, evidence-backed validation, and clean step-by-step commits. Use when work spans multiple files, has process risk, or needs strong project visibility.
Use this skill to execute medium-to-large engineering work with clear planning artifacts, granular execution tracking, and auditable outcomes.
Use this workflow when any of the following are true:
If the repository already has a planning convention, follow it.
If not, create a planning tree (example):
plans/<initiative>/
00-requirements.md
10-implementation-plan.md
20-task-board.yaml
30-compatibility-matrix.md # optional, for parity/migration work
40-test-strategy.md # optional, for complex validation
Create/update 00-requirements.md with:
For large features, add dedicated requirement deep dives in separate docs.
Use separate requirement docs when one feature has at least one of:
Name them predictably (examples):
01-feature-auth-migration.md02-feature-package-reproducibility.mdTemplate: references/requirements-deep-dive-template.md
Create/update 10-implementation-plan.md with:
Break work into fine-grained tasks. Prefer tasks small enough to complete and validate in one focused iteration.
Task granularity guidance:
Template: references/implementation-plan-template.md
Create/update 20-task-board.yaml with machine-readable task state.
Each task should include:
id, title, status, priority, depends_oncommand + result)hash + message)Rules:
Template: references/task-board-template.yaml
For each task, do this sequence:
in_progress.Whenever behavior changes:
At completion:
references/requirements-deep-dive-template.mdreferences/implementation-plan-template.mdreferences/task-board-template.yaml