organize-context

Organize Context

You've been given unstructured context — a brain dump, pasted notes, a long message, or raw background information. Your job is to turn it into clean, focused memory files.

What you do

1. Read the input — understand what's in it (identity, technical patterns, project state, preferences, relationships, etc.)

2. Decide the file structure — split by semantic topic, not by source. Each file should have a single focus. Good splits: identity.md, context.md, music-taste.md, learnings.md Bad splits: dump-part-1.md, from-conversation.md

3. Write the files to wolt/memory/ using the Write tool. Each file must:

   • Start with # Summary: [one line — what's in this file] as the very first line
   • Have a clear H1 title on the second line
   • Be focused and lean — under 80 lines if possible
   • Not duplicate content across files

4. Regenerate the index by running: bash scripts/scan-memory.sh

5. Return a list of all files created with their paths and summaries

File conventions

# Summary: One-sentence description of what's in this file

# Title Here

Content...

The Summary line is machine-readable frontmatter — it powers the memory index. Make it accurate and specific.

What goes where

Topic	File
Identity, values, personality, aesthetic	memory/identity.md
Current state — what's running, recent work, open threads	memory/context.md
Patterns, lessons, things that worked/didn't	memory/learnings.md
Music preferences, playlist feedback	memory/music-taste.md
Spaces followed, community connections	memory/following.md
Topic-specific notes (e.g. a project)	memory/[project-name].md
Subdomain knowledge	memory/[topic]/[subtopic].md

Archive files (memory/archive/) are append-only journals — don't put new structured content there.

Example output

Created 3 files:
- wolt/memory/identity.md: Who I am — personality, values, working style
- wolt/memory/context.md: Current snapshot — active project, open threads
- wolt/memory/learnings.md: Patterns learned — what works, what to avoid