Skip to content
Back to skills

Plan

ASecurity

Explore the codebase and produce an implementation plan. Use when the user runs /plan or wants a plan before building.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 22, 2026
testingtypescriptgobashgitapi

Works with

  • api

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned September 22, 2026

npx -y skills add jupes/agent-forge-harness --skill plan --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Plan?

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

Security grade badge for Plan
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/jupes-plan/badge)](https://www.skillsdirectory.com/skills/jupes-plan)

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: plan
description: Explore the codebase and produce an implementation plan. Use when the user runs /plan or wants a plan before building.
---

# /plan — Explore + Create Implementation Plan

Explores the codebase and produces a structured implementation plan. Optionally materializes Beads tasks.

## Usage
```
/plan add user settings page
/plan T-42                     # Plan for an existing task
/plan refactor auth module --tasks   # Also create Beads tasks from the plan
```

---

## Step 1 — Load Knowledge (mandatory)

Before any code exploration, load the relevant knowledge files:

```bash
cat knowledge/_shared.yaml
cat knowledge/repos/<repo>.yaml      # For each affected repo
```

If knowledge files are missing or stale (>50 commits behind HEAD):
- Warn: "knowledge/<repo>.yaml may be stale — consider running /sync-knowledge <repo>"
- Proceed anyway with a note that exploration results may supersede stale knowledge

---

## Step 1b — Planner mode *(short or vague prompts)*

If the user input is **only one to four sentences** with **no** Beads task id and **no** detailed spec:

1. Read `.claude/agents/planner.md` and adopt the **Planner** role **before** deep exploration.
2. Expand the prompt into a **product spec** (overview, numbered features / user stories, **non-goals**, testable signals). Save to `plans/drafts/<slug>-product-spec.md`.
3. Use that spec as the source of truth for exploration in Step 3 and the implementation plan in Step 4.

If the user already provided a detailed spec or `bd show` task context, skip this step.

---

## Step 2 — Reuse Verification

Search for existing implementations before planning anything new:

```bash
# Hooks
grep -rE "^export function use" src/ --include="*.ts" --include="*.tsx" -l

# Types and interfaces
grep -rE "^export (interface|type)" src/ --include="*.ts" -l

# Components
grep -rE "^export (default|function|const)" src/components/ --include="*.tsx" -l

# Utilities
grep -rE "^export (function|const|class)" src/utils/ --include="*.ts" -l
```

Note any existing implementations relevant to the plan.

---

## Step 3 — Explore

Use an Explore subagent (or explore directly) to map the codebase:

For small scope (single feature area):
- Read 5-10 key files relevant to the task
- Trace the data flow from entry point to the affected code

For large scope (cross-cutting concern):
- Spawn an Explore agent: "Explore <repo> and map all files related to <topic>. Focus on: <areas>. Report file paths, key functions, and integration points."
- Read the agent's findings, then read the specific files it identifies

**Goal**: understand the codebase well enough to write a complete, accurate plan with real file paths and function names.

---

## Step 4 — Write Plan

Write to `plans/drafts/<TASK-ID>.md` (create `plans/drafts/` if needed).
If no task ID: use a slug from the description.

### Plan format:

```markdown
# Plan: <TASK-ID> — <title>
Generated: <date>
Repo: <repo>

## Summary
<~100 tokens: what is being built, why, and the approach in plain language>

## Existing Code to Reuse
- `src/hooks/useAuth.ts` — auth state hook, extend it for settings
- `src/types/user.ts` — UserProfile type already has the needed fields

## Implementation Steps
1. Add settings schema to `src/types/user.ts`
   - Extend `UserProfile` with `settings: UserSettings`
   - Define `UserSettings` interface

2. Create `src/api/settings.ts`
   - `getSettings(userId: string): Promise<UserSettings>`
   - `updateSettings(userId: string, patch: Partial<UserSettings>): Promise<void>`

3. Create `src/hooks/useUserSettings.ts`
   - Wraps `getSettings` / `updateSettings`
   - Returns `{ settings, loading, error, update }`

4. Create `src/components/UserSettingsPage.tsx`
   - Form bound to `useUserSettings`
   - Save/cancel buttons, loading state, error display

5. Add route in `src/App.tsx`
   - `/settings` → `<UserSettingsPage />`

## Files to Modify
| File | Change |
|------|--------|
| `src/types/user.ts` | Add UserSettings interface |
| `src/App.tsx` | Add /settings route |

## Files to Create
| File | Purpose |
|------|---------|
| `src/api/settings.ts` | Settings API calls |
| `src/hooks/useUserSettings.ts` | Settings state hook |
| `src/components/UserSettingsPage.tsx` | Settings UI |
| `src/api/settings.test.ts` | API unit tests |
| `src/hooks/useUserSettings.test.ts` | Hook unit tests |

## Test Strategy
- Unit: `useUserSettings` — loading state, error state, successful update
- Unit: `settings.ts` API — mock fetch, verify request format
- Integration: settings page saves and persists across page refresh
- Edge: empty settings object, network error, validation failure

## Validation Commands
\`\`\`bash
bun run typecheck
bun test src/api/settings.test.ts
bun test src/hooks/useUserSettings.test.ts
\`\`\`

## Estimated Scope
- Files: 5 new, 2 modified
- Complexity: Medium
- Estimated time: 90 minutes

## Interfaces (if this plan produces shared APIs)
\`\`\`typescript
export function useUserSettings(userId: string): {
  settings: UserSettings | null;
  loading: boolean;
  error: Error | null;
  update: (patch: Partial<UserSettings>) => Promise<void>;
}
\`\`\`
```

---

## Step 5 — Present Plan

Output a summary to the user:
```
Plan ready: plans/drafts/<TASK-ID>.md

Summary: <2-sentence description>
Scope: <n> files (<n> new, <n> modified), ~<n> minutes
Complexity: Low | Medium | High

Key decisions:
- <decision 1 and why>
- <decision 2 and why>
```

---

## Step 6 — Materialize Beads *(with `--tasks` flag only)*

If `--tasks` was passed:

For each implementation step in the plan:
```bash
bd create \
  --type task \
  --title "<step description>" \
  --repo <repo> \
  --priority <value> \
  --acceptance "<acceptance criterion from the step>"
```

Pick **`<value>`** with `.claude/skills/beads-priority-assignment/SKILL.md` (critical-path or high-risk steps → **high** / **critical**; later optional steps → **medium** / **low**).

Add sequential dependencies (step 2 requires step 1):
```bash
bd dep add <step-2-id> --requires <step-1-id>
```

If there are 4+ tasks: create a parent feature:
```bash
bd create --type feature --title "<overall feature title>" --priority <value>
# then add parent relationship to each task
```

Use the same skill for the parent feature priority (often **medium** unless the plan is release-blocking).

Report: "Created N Beads tasks. Run /go to start work."

---

## Staleness Warning

After loading knowledge files, check commit distance:
```bash
git log --oneline knowledge/repos/<repo>.yaml -- | head -1
git rev-list --count HEAD   # compare against knowledge file's last commit
```

If >50 commits behind: `⚠ knowledge/<repo>.yaml is stale (>50 commits). Run /sync-knowledge <repo>.`

Files in this skill

  • SKILL.md6.6 KB
  • agents/openai.yaml217 B

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.