← Back to Plugins
Tools

Clawpound Knowledge

wonderchook By wonderchook 👁 7 views ▲ 0 votes

Knowledge work loop for OpenClaw: brainstorm, plan, review, execute, and compound learnings. Inspired by EveryInc/compound-knowledge-plugin.

GitHub

Install

openclaw plugins install -l

Configuration Example

{
  "plugins": {
    "entries": {
      "clawpound-knowledge": {
        "enabled": true,
        "config": {
          "knowledgeSubdir": "knowledge",
          "plansDir": "plans",
          "autoSuggestCompound": true
        }
      }
    }
  },
  "agents": {
    "defaults": {
      "tools": {
        "enable": ["clawpound-knowledge"]
      }
    }
  }
}

README

# Clawpound Knowledge

Knowledge work loop for [OpenClaw](https://github.com/openclaw/openclaw): brainstorm, plan, review, execute, and compound learnings into durable memory.

Inspired by [EveryInc/compound-knowledge-plugin](https://github.com/EveryInc/compound-knowledge-plugin) by Austin Tedesco at [Every](https://every.to). Adapted for OpenClaw's plugin system and memory architecture.

## What It Does

Clawpound Knowledge provides six workflow skills that form a compounding loop:

1. `/kw:brainstorm` -- Brain dump capture, extract decisions/questions/constraints, search knowledge base
2. `/kw:plan` -- Research past work, structure plans grounded in data and past learnings
3. `/kw:confidence` -- Epistemic gut-check before proceeding
4. `/kw:review` -- Parallel strategic alignment and data accuracy review
5. `/kw:work` -- Execute plans batch-by-batch with progress tracking
6. `/kw:compound` -- Extract and save learnings so the next cycle starts smarter

Each cycle produces learnings saved to `memory/knowledge/`, which OpenClaw's `memory_search` automatically indexes. Future planning cycles find these learnings, closing the loop.

## Installation

```bash
openclaw plugins install -l /path/to/clawpound-knowledge
```

Or via npm (once published):

```bash
openclaw plugins install @wonderchook/clawpound-knowledge
```

## Configuration

Enable the tools in your OpenClaw config:

```json
{
  "plugins": {
    "entries": {
      "clawpound-knowledge": {
        "enabled": true,
        "config": {
          "knowledgeSubdir": "knowledge",
          "plansDir": "plans",
          "autoSuggestCompound": true
        }
      }
    }
  },
  "agents": {
    "defaults": {
      "tools": {
        "enable": ["clawpound-knowledge"]
      }
    }
  }
}
```

## Memory Integration

Knowledge entries are stored in `memory/knowledge/` within the OpenClaw workspace directory. This means:

- `memory_search` automatically indexes and finds knowledge entries via hybrid BM25 + vector search
- Knowledge entries are "evergreen" (no temporal decay since filenames aren't date-based)
- The `kw_knowledge_search` tool adds structured search by tags, type, and confidence

Plans are stored in `plans/` at the workspace root -- they're working documents, not durable knowledge.

## Tools

Three opt-in tools complement the skills:

- `kw_knowledge_save` -- Save learnings with YAML frontmatter to `memory/knowledge/`
- `kw_knowledge_search` -- Search knowledge entries by tags, type, or content
- `kw_knowledge_check_stale` -- Find entries that conflict with or are superseded by new learnings

## License

MIT -- see [LICENSE](LICENSE).

## Credits

This project is inspired by the [compound-knowledge-plugin](https://github.com/EveryInc/compound-knowledge-plugin) created by Austin Tedesco at [Every, Inc](https://every.to). The original plugin provides the same knowledge work loop for Claude Code. Clawpound Knowledge adapts the concept for OpenClaw's plugin system, integrating with OpenClaw's memory architecture.
tools

Comments

Sign in to leave a comment

Loading comments...