← Back to Skills Marketplace
192
Downloads
1
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install metaclaw
Description
Generischer Skill für Memory-Management, Hybrid Retrieval und Skill Evolution mit PluginEval Integration
Usage Guidance
This skill appears coherent with its stated purpose, but review and take these precautions before installing or enabling automatic runs:
- Inspect and run in an isolated environment (use the advised .venv-metaclaw) so Python packages and model downloads are separated from your system.
- The scripts read and write files under your workspace (MEMORY.md, memory/*.md, SESSION-STATE.md, .session-count). Back up important data before running consolidation/auto-fix steps.
- memory-extract sends content to a local Ollama endpoint (http://localhost:11434). Confirm you run a trusted local LLM; do not point OLLAMA_HOST (if you modify code) to an untrusted remote service or put sensitive data into prompts sent to external hosts.
- skill-eval's auto-fix can modify files when invoked with --allow-write; run first with read-only flags (--layer1, --auto-fix without --allow-write) to preview changes. Keep backups and review diffs before applying fixes.
- There are minor path/name inconsistencies in docs vs. scripts (e.g., references to memory-digest-enhanced.sh vs memory-digest.sh and varying script paths). Verify the correct paths before installing crontab or heartbeat hooks.
If you want higher confidence, provide answers or artifacts for these points and I can re-evaluate with higher confidence:
- Do you plan to run a local Ollama or other LLM? If not, which extraction fallback do you intend to use?
- Will you allow CLI '--allow-write' or cron/heartbeat integration (automated writes)?
- Do you want me to scan specific files in your workspace to see what this skill would read/process?
Capability Analysis
Type: OpenClaw Skill
Name: metaclaw
Version: 1.0.0
The MetaClaw skill bundle is a legitimate toolset for advanced memory management and skill quality maintenance within the OpenClaw ecosystem. It features hybrid retrieval (semantic and FTS) using LanceDB, automated memory extraction via a local Ollama instance (memory-extract.py), and a static analysis tool (skill-eval.py) designed to improve skill metadata. All operations are confined to the local workspace, and the 'auto-fix' functionality is protected by explicit safety flags. No indicators of data exfiltration, malicious execution, or harmful prompt injection were detected.
Capability Assessment
Purpose & Capability
Name/description (MetaClaw: memory management, hybrid retrieval, PluginEval) align with the included scripts: memory vector search, memory extraction, session counter, digest/consolidation, and a skill-evaluator. Declared requirements (none) are plausible because defaults exist, and the SKILL.md explicitly instructs creating a venv and installing lancedb and sentence-transformers.
Instruction Scope
Runtime instructions and scripts operate on the user's OpenClaw workspace (MEMORY.md, memory/*.md, SESSION-STATE.md, .session-count) and can read/parse those files — this matches the stated purpose. The memory-extract script may call local commands (e.g., lcm_describe) and sends prompts to a local Ollama API (http://localhost:11434) for LLM-based extraction. The skill-eval tool can modify files when run with --allow-write. These behaviors are expected for the described functionality but warrant review because they read/write local files and may invoke external tools on the host.
Install Mechanism
No formal install spec in registry (instruction-only), but SKILL.md/README instruct creating a Python venv and pip installing lancedb and sentence-transformers. That is proportionate to the code which imports those packages. Be aware sentence-transformers/models may download model weights on first run (network activity).
Credentials
The skill does not declare required env vars but the code honors WORKSPACE (defaulting to ~/.openclaw/workspace), OLLAMA_MODEL (optional, default provided), and OLLAMA_TIMEOUT. These are reasonable for a local extraction pipeline. No cloud/AWS/third-party API credentials are requested. A caveat: skill-eval's --allow-write gives the script permission to modify files you point it at; exercise caution before allowing writes.
Persistence & Privilege
Flags show no forced persistence (always:false) and model invocation is permitted (default). The skill writes only within the workspace and skill directories by default (e.g., session-count, backups, .last-digest). The ability to modify files is gated behind explicit CLI flags (--allow-write) which is appropriate; still, cron instructions and Heartbeat integration can make it run periodically if the user configures them.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install metaclaw - After installation, invoke the skill by name or use
/metaclaw - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release - Hybrid Retrieval, Memory Extraction, Session Counter, PluginEval Integration
Metadata
Frequently Asked Questions
What is Metaclaw?
Generischer Skill für Memory-Management, Hybrid Retrieval und Skill Evolution mit PluginEval Integration. It is an AI Agent Skill for Claude Code / OpenClaw, with 192 downloads so far.
How do I install Metaclaw?
Run "/install metaclaw" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Metaclaw free?
Yes, Metaclaw is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Metaclaw support?
Metaclaw is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Metaclaw?
It is built and maintained by Donmeusi (@donmeusi); the current version is v1.0.0.
More Skills