Skill: capture-idea

Trigger

Use this skill when the user says:

• "I have an idea"
• "capture idea"
• "new idea"
• invokes /capture-idea

Purpose

Capture a raw idea quickly into docs/planning/ideas/ with a date-prefixed filename and YAML frontmatter. Ideas are stage 1 of the planning pipeline: captured → evaluated → promoted to PRD.

Output

Creates: docs/planning/ideas/YYYY-MM-slug.md

Workflow

Step 1 — Take the idea as-is

Accept whatever the user provides — a single sentence is enough. Do not ask clarifying questions upfront.

Step 2 — Query the codebase

Before writing the file, search the codebase to understand what the idea refers to:

• Use Glob/Grep to find relevant files, components, schemas, or routes mentioned or implied by the idea
• Read any key files needed to understand current state
• Use this context to fill in the sections below with real, grounded content

Only ask the user a question if the idea is so ambiguous that you cannot make a reasonable inference about what it refers to in this codebase (e.g., a single word with no context). One targeted question maximum.

Step 3 — Infer metadata

From the description and codebase context, determine:

• slug: kebab-case from the title (e.g., "Add llms.txt file" → add-llms-txt-file)
• domain: one of seo | engineering | content | design | infrastructure
• source: one of analytics | competitor-research | internal | seo-data
• priority: default unset unless the user signals urgency
• status: always captured on creation

Step 4 — Create the file

Filename format: YYYY-MM-slug.md using today's date for YYYY-MM.

Use the template at docs/planning/templates/idea.md. Populate all frontmatter fields. Write concise content under each section heading based on what you found in the codebase — do not leave sections as empty placeholders. Leave prd: "" blank (filled when promoted).

Step 5 — Update the index

After creating the file, always update docs/planning/ideas/_index.md:

• Read all .md files in docs/planning/ideas/ (excluding _index.md)
• Parse each file's frontmatter to extract title, status, priority, and domain
• Rewrite _index.md with:
  • Updated updated date in frontmatter
  • A summary table with counts per status
  • An ideas table listing all entries (title, status, priority, domain, filename link)
• If _index.md does not exist, create it using this structure

Step 6 — Confirm

Tell the user the file was created at docs/planning/ideas/YYYY-MM-slug.md and ask if they want to immediately evaluate it (status: evaluating) or leave it as captured for later triage.

Frontmatter Schema

---
title: ''
status: captured
priority: unset
source: internal
captured: 'YYYY-MM-DD'
domain: seo
prd: ''
tags: []
---

Valid status values: captured | evaluating | accepted | rejected | deferred Valid priority values: critical | high | medium | low | unset Valid source values: analytics | competitor-research | internal | seo-data Valid domain values: seo | engineering | content | design | infrastructure

Template Sections

## Problem / Opportunity

## Context

## Rough Scope

## Success Signal

## Open Questions

## References

Notes

• Ideas directory: docs/planning/ideas/
• Template: docs/planning/templates/idea.md
• Date prefix uses YYYY-MM (year-month) not full date — ideas are a chronological stream