Skip to content
Back to skills

Planning Small Features

ASecurity

Planning the lightweight 4-gate Small Track pre-dev workflow (research, PRD, TRD, plan) with human approval and state tracking at each gate. Use for features under 2 days that reuse existing patterns and add no new dependencies, data models, or services. Skip for larger or complex features — use ring:planning-large-features instead. Plans only — no edits.

  • 215 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 6, 2026
developmentgofrontendbackend

Security analysis

A100/100

Scanned September 6, 2026

npx -y skills add LerianStudio/ring --skill planning-small-features --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Planning Small Features?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Planning Small Features
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/lerianstudio-planning-small-features/badge)](https://www.skillsdirectory.com/skills/lerianstudio-planning-small-features)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: ring:planning-small-features
description: "Planning the lightweight 4-gate Small Track pre-dev workflow (research, PRD, TRD, plan) with human approval and state tracking at each gate. Use for features under 2 days that reuse existing patterns and add no new dependencies, data models, or services. Skip for larger or complex features — use ring:planning-large-features instead. Plans only — no edits."
---

# Small Track Pre-Dev Workflow (4 Gates)

## When to use

- Feature takes <2 days to implement
- Uses existing architecture patterns
- Doesn't add new external dependencies
- Doesn't create new data models/entities
- Doesn't require multi-service integration
- Can be completed by a single developer

## Skip when

- Feature is complex (>=2 days) - use ring:planning-large-features instead
- Adds new dependencies, data models, or architecture patterns

## Sequence

**Runs before:** ring:running-dev-cycle, ring:executing-plans

## Related

**Complementary:** ring:planning-large-features, ring:creating-worktrees, ring:product-designer + ring:validating-ux-completeness (standalone UX step, recommended when feature has UI)
**Skills orchestrated:**
- ring:researching-features
- ring:writing-prds
- ring:writing-trds
- ring:writing-plans


Running the **Small Track** pre-development workflow for features that take <2 days, use existing patterns, add no new external dependencies, create no new data models, require no multi-service integration, and can be completed by a single developer.

For complex features (any of the above false), use `ring:planning-large-features` instead.

## Gate Map

| Gate | Skill | Output |
|------|-------|--------|
| 0 | ring:researching-features | research.md |
| 1 | ring:writing-prds | prd.md |
| 2 | ring:writing-trds | trd.md |
| 3 | ring:writing-plans | plan.md |

All artifacts saved to: `docs/pre-dev/<feature-name>/`

## Step 1: Gather Feature Name

AskUserQuestion: "What is the name of your feature?" (kebab-case, e.g., "user-logout", "email-validation")

## Step 2: Topology Discovery (MANDATORY)

Execute topology discovery per [shared-patterns/topology-discovery.md](../shared-patterns/topology-discovery.md). Store as `TopologyConfig` for all subsequent gates.

## Step 3: Gather Feature-Specific Inputs

**Q2 (CONDITIONAL):** Auth requirements — auto-detect from `go.mod` (`lib-auth` present → skip). Options: None, User only, User + permissions, Service-to-service, Full.

**Q3 (CONDITIONAL):** License requirements — auto-detect from `go.mod` (`lib-license-go` present → skip). Options: No, Yes.

**Q4 (MANDATORY):** Has UI? Options: Yes, No. Always ask — do not assume.

**Q5 (if Q4=Yes):** UI component library — auto-detect from package.json, confirm with user.

**Q6 (if Q4=Yes):** Styling approach — auto-detect from package.json, confirm with user.

## Step 4: Execute Gates Sequentially

All gates (0-3) always run. Human approval required at each gate before proceeding.

**Gate 3 (Plan):** invoke `ring:writing-plans` with trd.md as the spec input (no feature-map on Small Track), passing `TopologyConfig`. Output path: `docs/pre-dev/{feature}/plan.md` (overrides the writing-plans default). plan.md is always a SINGLE document per feature. **Topology clause:** when `TopologyConfig` structure is monorepo or multi-repo, each epic carries one line `**Target:** backend | frontend | infra` (placed right before `**Status:**`); for multi-repo, the orchestrator copies plan.md into each repo and the local dev-cycle executes only epics whose Target matches that repo. No per-module plan splits.

**Standalone UX step (if Q4=Yes):** after Gate 1 approval, RECOMMEND running `ring:product-designer` + `ring:validating-ux-completeness` before Gate 2 (TRD). It is optional, not a gate, and not tracked in workflow-state.json. If design-validation.md exists when Gate 2 runs, the TRD honors its verdict; if absent, proceed and note the UX risk.

## Gate Progress Tracking

Save state to `docs/pre-dev/{feature}/workflow-state.json`:
```json
{
  "track": "small",
  "feature": "{feature-name}",
  "currentGate": 0,
  "gates": {"0": "PENDING", "1": "PENDING", "2": "PENDING", "3": "PENDING"},
  "topology": {},
  "inputs": {"hasUI": false, "authRequired": false, "licenseRequired": false, "uiLibrary": null, "styling": null}
}
```

## Execution Mode

AskUserQuestion at start: "Execution mode?" Options: Automatic (pause only on failure), Manual (checkpoint after each gate).

## Completion

After Gate 3 approved: `docs/pre-dev/{feature}/plan.md` is the single execution document. Execute with `ring:running-dev-cycle` (subagent orchestration) or `ring:executing-plans` (inline).

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.