Skip to content

README.md

Latest commit

 

History

History
46 lines (34 loc) · 3.82 KB

README.md

File metadata and controls

46 lines (34 loc) · 3.82 KB
Raw

Loop templates

Copy-and-edit starting points for everyday Loop work. Each .loop here is a heavily commented, best-practice skeleton for a common job — fixing a bug, shipping a feature, driving CI to green, delivering a spec story by story. They are written to be cloned into your repo and customized: every line that's specific to your project is marked # TODO (test commands, file paths, try ceilings). All of them parse cleanly, so show/run work the moment you copy them. Don't run a template as-is — fill in the TODOs first.

Pick a template

Spec-driven (a whole app / a written backlog)

Template When to use Shape
greenfield-app.loop Build an app from nothing, A to Z. flow: discover.loop → design.loop → for each story in "sprint.yaml" → story-template.loop
load-spec.loop You already have a spec + backlog; deliver it story by story. flow: for each story in "sprint.yaml" → story-template.loop

Change (build / fix / restructure)

Template When to use Shape
feature.loop Ship one feature with regression + security + human review. pipeline (fail-fast): build → regression → security → 👤 review
brownfield-feature.loop Add a feature to an existing codebase without breaking it. pipeline (fail-fast, file-level git: branch/commit/no-push): build → regression (no-regression guarantee) → security → 👤 review
bugfix.loop Fix one bug, proven by a named test. single loop: reproduce → fix → verify; ↺ reflect-on-fail, ⛔ after 6 tries
refactor.loop Improve structure with behavior unchanged. single loop: full suite green = no-regression proof; no-new-deps gate; 👤 review; ⛔ after 6

Quality gates (drive existing checks to green)

Template When to use Shape
cicd-check.loop Make every CI check pass locally. pipeline (fail-fast, git: branch/commit/no-push): lint → typecheck → test → build
security.loop Run a security pass before shipping. pipeline (fail-fast): sast (semgrep) → deps (npm audit) → 👤 secrets (gitleaks)
clean-architecture.loop Enforce architecture boundaries — deps point inward, no layer leaks. single loop: done when a boundary checker passes (or a review skill scores); ↺ reflect = invert/move, don't relax the rule; 👤 review; ⛔ after 6
test-coverage.loop Raise coverage to a threshold with meaningful tests. single loop: done when coverage gate passes; ↺ reflect-on-fail; ⛔ after 6
review-diff.loop Review and clean the current branch diff. single loop: done when the skill "code-review" approves; 👤 human signs off last; ⛔ after 4

How to use

  1. Copy the .loop you want into your repo and rename it.
  2. Edit every # TODO line — your real test commands, file paths, and try counts. (The default gate strings like pnpm test or npm test are placeholders; swap in yours.)
  3. Eyeball the shape, then run it:
    • /loopflow run <file> inside Claude Code, or
    • loop-run run <file> headless. Run show <file> first to sanity-check the flow before executing.

Supporting files

  • sprint.yaml and plan.md are starter data, not loops — a small auth-style backlog/spec to replace with your own. load-spec.loop and greenfield-app.loop iterate sprint.yaml; plan.md is the human-readable spec you keep in sync with it.
  • story-template.loop is the per-story checklist (implement → security → 👤 manual) shared by both load-spec.loop and greenfield-app.loop — author it once, drive it with the spec.
  • discover.loop and design.loop are the first two stages of greenfield-app.loop (interview → sprint.yaml, then sprint.yaml → design.md), each with its own artifact-check gate and human review.