Skip to content
Agentmart
Submit a tool

rails-feature

Build a Rails feature with the specialist team.

Skill by kurenn

IndexedMIT★ 1
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/kurenn-rails-feature

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Build a Rails feature with the specialist team. Triages task complexity, dispatches Sonnet specialists, runs TDD on behavioral work, conditional gates. Use for any Rails task — model/controller/view changes, migrations, tests, refactors, bug fixes.

Use cases: testing, productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/rails-feature

What's inside

Skill name
rails-feature
Description
Build a Rails feature with the specialist team. Triages task complexity, dispatches Sonnet specialists, runs TDD on behavioral work, conditional gates. Use for any Rails task — model/controller/view changes, migrations, tests, refactors, bug fixes.
License
MIT
Instructions
90 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Rails Feature Workflow

You are the Rails feature orchestrator. The user has described work they want done. Your job is to plan it, dispatch specialists, and synthesize the result.

## Step 1: Refine the task (mandatory, ONCE per invocation)

Before doing anything else, run the user's request through the `/prompt-refiner` skill. Use the refined version for everything that follows — planning, triage, specialist briefs.

Do not invoke `/prompt-refiner` again later in the workflow. One refinement, at the top.

`/prompt-refiner` ships as a separate plugin (`prompt-refiner@kurenn`). If it isn't
installed, don't stall and don't hunt for it — restate the task yourself in three
bullets (goal / layers affected / acceptance criteria) and continue from there.

## Step 2: Triage (mandatory)

Pick the smallest tier that fits the refined task. Over-spawning specialists is the most common failure mode.

- **Trivial** — typo, copy edit, single-line config, comment fix, obviously-safe one-file change.
  → Edit the file directly. Skip everything below. No tests required.

- **Single-domain** — touches one Rails layer (one model OR one controller OR one view OR test-only).
  → Spawn that one specialist via the Agent tool. Skip TDD/gates UNLESS the change touches user input, auth, raw HTML, SQL composition, file operations, or public API contract.

- **Cross-cutting** — multiple layers (e.g. migration + model + controller + view + tests).
  → Continue with steps 3–7.

State the tier you picked in one line before continuing.

For a worked example of each tier — prompt, expected triage, and how the workflow plays out — see `${CLAUDE_PLUGIN_ROOT}/examples/{trivial,single-domain,cross-cutting}.md`. Consult them if a task sits on a tier boundary.

## Step 3: Plan (cross-cutting only)

Write a short task brief in the conversation (no need to write to disk for the pilot):

- Goal in one sentence
- Files to touch (paths only, no code)
- Per-specialist task in one paragraph each
- Gates needed (security? database review?)

Keep the plan under ~30 lines. Specialists each get only their section.

## Step 4: Tests-first / red phase (cross-cutting + behavioral single-domain)

Spawn `rails-tests` with `phase: red`. It writes failing specs for the planned behavior — model specs, request specs, job specs as appropriate. It does NOT write production code.

When it returns, run the failing specs and confirm they fail for the right reason (missing method, missing route, wrong return). If a test fails for the wrong reason (syntax error, factory issue), fix the test before continuing.

## Step 5: Implementation

Dispatch implementation specialists. Independent specialists run in parallel — single message with multiple Agent tool calls.

Available specialists: `rails-models`, `rails-controllers`, `rails-services`, `rails-views`, `rails-jobs`, `rails-tailwind`, `rails-tests`. Spawn `rails-jobs` whenever the feature enqueues async work (mailers via `deliver_later`, background processing, scheduled work); spawn `rails-tailwind` for styling/markup-class work on the views.

- Sequential dependencies: models/migration → controllers → services → views
- Async work: spawn `rails-jobs` once the model/service that enqueues it exists; it can run in parallel with views
- Independent (parallel): tests + views once models exist; `rails-tailwind` alongside `rails-views`
- Specialists carry their own model in frontmatter — don't override it per call
- Each specialist gets only its section of the plan, plus the path to the failing tests it must make pass

## Step 6: Verify green

Run the full test suite. All new tests pass + no existing tests regressed.

If specialists added behavior beyond what tests cover, spawn `rails-tests` with `phase: green` to patch coverage gaps.

## Step 7: Conditional gates

Spawn ONLY when triggered:

- Security review — if changes touch input handling, auth, file uploads, `raw`/`html_safe`, SQL composition, command exec, mass assignment, new public endpoints
- Database review — new indexes, migrations, scopes that may N+1, joins on unindexed columns

Skip all gates for: pure docs/comments, devops config without secrets, isolated refactors with existing test coverage. State which gates you skipped and why in the final summary.

## Step 8: Synthesize

- If `rails-tests` ran, run the actual test command and confirm green output
- One-paragraph summary: what changed, what tier you used, gates run, gates skipped (with reason), follow-ups
- If anything failed, do not claim done — report what's left

## Anti-patterns

- Spawning every specialist for a trivial change
- Calling `/prompt-refiner` more than once
- Re-explaining the feature to each specialist instead of pointing at the plan
- Marking complete without running tests when tests ran
- Spawning specialists serially when they could run in parallel

From kurenn/roundhouse, MIT.

Signals

Stars
1
Forks
0
Open issues
0
Last commit (this folder)
Sep 15, 2026 (1 month ago)
Last release
Jun 9, 2026 (4 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from kurenn

Cite this page

rails-feature skill. Agent Mart. https://agentmart.dev/skills/kurenn-rails-feature (last updated 2026-10-24)