q404365631/coco/brain-export

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Auto-generate a slim CLAUDE.local.md from brain DB. Preserves manual behavior rules. Produces ~1,200 token auto-generated context section with key people, systems, decisions, and project status.

Install

npx vskill@latest install q404365631/coco/brain-export

SKILL.md

/brain-export — CLAUDE.local.md Auto-Generator

Generates or refreshes the <!-- AUTO-GENERATED: brain-export --> section in the project's CLAUDE.local.md file from the brain SQLite DB. Protects all manually written content above and below the sentinels.

Quick Reference

bash
BRAIN="python3 ~/.claude/skills/brain/scripts/brain/brain_cli.py"
# Basic export (writes CLAUDE.local.md next to project_brain.db)
$BRAIN export <project-slug>
# Export with explicit output path
$BRAIN export <project-slug> --output /path/to/CLAUDE.local.md
# Export with explicit DB path
$BRAIN --db /path/to/project_brain.db export <project-slug>

Procedure

Step 1 — Verify brain DB exists
bash
BRAIN="python3 ~/.claude/skills/brain/scripts/brain/brain_cli.py"
$BRAIN info

If the DB is missing, tell the user to run /brain-init first.

Step 2 — Run the export
bash
$BRAIN export <project-slug>

The exporter will:

  1. Read any existing CLAUDE.local.md at the output path
  2. Preserve everything above <!-- AUTO-GENERATED: brain-export -->
  3. Preserve everything below <!-- END AUTO-GENERATED -->
  4. If no sentinels found, append the auto section at the bottom (safe first run)
  5. Regenerate the auto section from brain DB
  6. Write atomically (tmp file + rename)
Step 3 — Show the user what was generated

Parse the JSON output and display:

brain:export complete
─────────────────────────────────────────
Output: /path/to/CLAUDE.local.md
Project slug: <slug>
Auto section: ~NNN tokens (within 1,200 token budget: YES/NO)
Total file: ~NNN tokens
─────────────────────────────────────────
Sections generated:
- Project Identity (if ~/.mempalace/identity.txt exists)
- Last Session (most recent event + decision)
- Key People (up to 8 person entities)
- Key Systems (up to 8 system/module entities)
- Recent Decisions (last 4 decisions)
- Project Status (entity/decision/event counts + last sync date)
Step 4 — First-run: offer to migrate feedback memories

If this is the first export AND the project memory directory contains files with type: feedback frontmatter, offer to migrate them into behavior rules:

Found N feedback memory files that could become behavior rules in CLAUDE.local.md:
- feedback_html_docs_with_tailwind.md → "Documents as Tailwind HTML not Markdown"
Add these as a ## Behavior Rules section below the auto-generated block? [y/N]

To migrate programmatically:

python
from brain.exporter import migrate_feedback_memories
from pathlib import Path
memory_dir = Path("~/.claude/projects/<project-slug-hash>/memory").expanduser()
rules = migrate_feedback_memories(memory_dir)
for rule in rules:
print(rule)

Then append the rules as a ## Behavior Rules section below <!-- END AUTO-GENERATED --> in CLAUDE.local.md.

Auto-Section Format

markdown
<!-- AUTO-GENERATED: brain-export -->
<!-- Generated: 2026-04-09T12:00:00Z | Brain DB: 42 entities | MemPalace: 7 drawers -->
## Project Identity
<contents of ~/.mempalace/identity.txt if it exists>
## Last Session
-Most recent event title (YYYY-MM-DD)
-Most recent decision text (YYYY-MM-DD)
## Key People
| Name | Role | Email |
|------|------|-------|
| Alice Smith | Product Manager | alice@example.com |
## Key Systems
-**SystemA** — Description from metadata
-**ModuleB**
## Recent Decisions
-[2026-04-08] Decision text (truncated to 120 chars)
## Project Status
-Entities: 42 | Decisions: 12 | Events: 8
-Last brain sync: 2026-04-09
-Query: `/coco brain` or `/brain context {slug}` for full context
<!-- END AUTO-GENERATED -->

Sentinel Contract

SentinelPositionMeaning
<!-- AUTO-GENERATED: brain-export -->Start of auto blockEverything below is managed by exporter
<!-- END AUTO-GENERATED -->End of auto blockEverything below is protected manual content

Never manually edit between the sentinels — changes will be overwritten on next export. Safe to edit anything above START or below END at any time.

Token Budget

The auto section is capped at 1,200 tokens (~4,800 characters). If the generated content exceeds this, the body is truncated. The header comment and sentinels are always preserved intact.

When to Run

  • After every /brain-update session flush (brain DB has new data)
  • After /brain-init (first time setup)
  • After /brain-rescan (bulk entity import)
  • When starting work on a project for the first time in a new session (ensure context is fresh)

Trigger Phrases

  • "export brain to CLAUDE.local.md"
  • "refresh project context file"
  • "generate CLAUDE.local"
  • "update auto-generated section"
  • "brain export <slug>"

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/q404365631/coco/brain-export/badge)](https://verified-skill.com/skills/q404365631/coco/brain-export)

Something wrong with this skill or its page? Tell us.