This repository has been archived on 2026-10-25. You can view files and clone it. You cannot open issues or pull requests or push a commit.
Files
Dragon 66dd1cee19 v2: trim SKILL.md, add tiers/re-anchor/archives/user guide
Changes:
- SKILL.md cut to ~1400 tokens (from ~2500+)
- Schema examples moved to references/
- Project tiers: Nano/Small/Large
- Mid-session re-anchor rule (every 15+ tool calls)
- SESSIONS + ERRORS + LEARNINGS archives
- Wrong-turn report format with error logging
- Collapsed session end to 5 steps + self-reflection
- Pending errors check at session start
- DEPS.md template for dependency tracking
- learnings-guide.md reference doc
- GUIDE.md: comprehensive user guide (shared with SmallClaw)
- README.md rewritten with comparison table
2026-04-02 20:02:44 +00:00

2.2 KiB
Raw Permalink Blame History

SmallClawPlus

Structured development workflow for AI models with self-improving learnings and dependency tracking. Extends SmallClaw.

Works with medium models (30B+). Great for large ones too.

What it adds over SmallClaw

Feature SmallClaw SmallClawPlus
Brain files (STATUS, CONTEXT, SESSIONS) ✅ ✅
Project tiers (Nano/Small) ✅ ✅ + Large tier
Mid-session re-anchor ✅ ✅
Error capture (.learnings/ERRORS.md) ❌ ✅
Pattern learning (.learnings/LEARNINGS.md) ❌ ✅
Learning promotion to ARCH.md ❌ ✅
Dependency tracking (DEPS.md) ❌ ✅
Self-reflection at session end ❌ ✅
Feature request tracking ❌ ✅

When to use which

  • SmallClaw — small models (8B-30B), simple projects, minimal overhead
  • SmallClawPlus — medium+ models (30B+), projects that benefit from persistent error memory and dependency awareness

Project Tiers

Tier Source files Brain files Mandatory tokens
Nano <5 STATUS + CONTEXT + SESSIONS ~130
Small 5–20 + PLAN + ARCH + GLOSSARY + CHANGELOG + DEPS + .learnings/ ~700
Large 20+ Full stack with module protocol ~700

File Structure

your-project/
├── brain/
│   ├── STATUS.md
│   ├── CONTEXT.md
│   ├── SESSIONS.md
│   ├── ARCH.md
│   ├── PLAN.md
│   ├── DEPS.md
│   ├── GLOSSARY.md
│   ├── CHANGELOG.md
│   └── .learnings/
│       ├── ERRORS.md
│       ├── LEARNINGS.md
│       └── FEATURE_REQUESTS.md
└── ...source files...

Quick Start

Install the skill:

mkdir -p /path/to/openclaw/skills/dev-project
cp SKILL.md /path/to/openclaw/skills/dev-project/SKILL.md

Then start a conversation:

"Start a new project — TailScan. Python web app, ~12 files. Gitea at git.rewr.ca."

User Guide

Read GUIDE.md — teaches you how to communicate effectively with the model using this system. Shared with SmallClaw.

See Also

  • SmallClaw — lighter version without learnings
  • GUIDE.md — how to talk to the model effectively