← Back to Skills Marketplace
BEE Belief Extraction Engine
by
Kartik Vashisth
· GitHub ↗
· v0.1.1
473
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install openclaw-bee
Description
Install and configure BEE — the Belief Extraction Engine for OpenClaw. Gives agents persistent structured memory across sessions. Auto-extracts beliefs at se...
Usage Guidance
Before installing: 1) Inspect the package source on the referenced GitHub repo and the npm package page; check for unexpected postinstall scripts or obfuscated code. 2) Verify the package author and recent release history (who published @skysphere-labs/openclaw-bee). 3) Understand how your OpenClaw installation supplies LLM credentials (Anthropic API key or other provider) — the skill uses an extraction model but does not declare required credentials. 4) Consider installing first in an isolated environment/container or non-production agent. 5) Back up ~/.openclaw/openclaw.json and any existing state, and review file permissions for the DB path; determine whether the DB should be encrypted or have a retention policy for sensitive beliefs. 6) If you proceed, review runtime logs (enable debug) and periodically inspect the SQLite DB to ensure only expected data is being stored. If you are not comfortable auditing the package or exposing sensitive conversation data, do not install.
Capability Analysis
Type: OpenClaw Skill
Name: openclaw-bee
Version: 0.1.1
The OpenClaw Bee skill bundle is designed to provide persistent, structured memory for OpenClaw agents. The `SKILL.md` file outlines installation steps involving `npm` to install the `@skysphere-labs/openclaw-bee` package, configuration of `~/.openclaw/openclaw.json`, and restarting the OpenClaw gateway. Verification steps include querying the agent or directly accessing a local SQLite database (`vector.db`). All instructions and actions described are consistent with the stated purpose, involve standard package management and configuration, and show no evidence of data exfiltration, unauthorized network activity, persistence mechanisms beyond the intended extension, or malicious prompt injection attempts against the agent.
Capability Assessment
Purpose & Capability
Name/description describe a persistent belief store and the SKILL.md instructs installing an npm package, configuring a local SQLite DB, and enabling the extension — these are coherent. However the skill relies on an LLM extraction model (anthropic/claude-haiku-4-5) yet the SKILL.md and registry metadata do not declare any required credentials or explain how model credentials are provided; this is an unexplained omission.
Instruction Scope
Instructions are specific: npm install -g the package, edit ~/.openclaw/openclaw.json to enable BEE and set dbPath/agentId, then restart the gateway. The runtime steps reference only the OpenClaw config and a SQLite DB path, which match the stated purpose and do not instruct broad system or network reconnaissance. They do, however, include a direct sqlite3 command to inspect the DB (expected for this purpose).
Install Mechanism
Although there is no automated install spec in the registry, the SKILL.md directs users to run 'npm install -g' (or install from GitHub). Installing a global npm package can execute arbitrary postinstall scripts and will place code on disk; this is a moderate-risk install mechanism. The SKILL.md does not advise verifying the package or reviewing its postinstall behavior.
Credentials
The skill requests no environment variables in registry metadata, but its default configuration uses an external LLM provider (Anthropic). Calling that model normally requires API credentials (or provider configuration) which are not described or declared. Additionally, the skill persists potentially sensitive conversation data indefinitely to a local SQLite DB (vector.db) with no mention of encryption, access controls, or retention — this increases data exposure risk.
Persistence & Privilege
The skill is not always: true and does not request elevated platform privileges. It modifies the user's OpenClaw extension config (~/.openclaw/openclaw.json) and writes a SQLite DB to ~/.openclaw/workspace/state — these are within the extension's stated scope but do grant persistent local storage and long-term access to extracted beliefs, which may contain sensitive content.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install openclaw-bee - After installation, invoke the skill by name or use
/openclaw-bee - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.1.1
- Initial public release of openclaw-bee (v0.1.1).
- Provides BEE, the Belief Extraction Engine, for persistent, structured agent memory across OpenClaw sessions.
- Supports namespace-based belief scoping, automated belief extraction/deduplication, and session context recall for agents.
- Offers quick install via npm or GitHub and detailed configuration instructions.
- Beliefs are stored in a persistent SQLite database; includes methods for verification and multi-agent setup.
Metadata
Frequently Asked Questions
What is BEE Belief Extraction Engine?
Install and configure BEE — the Belief Extraction Engine for OpenClaw. Gives agents persistent structured memory across sessions. Auto-extracts beliefs at se... It is an AI Agent Skill for Claude Code / OpenClaw, with 473 downloads so far.
How do I install BEE Belief Extraction Engine?
Run "/install openclaw-bee" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is BEE Belief Extraction Engine free?
Yes, BEE Belief Extraction Engine is completely free (open-source). You can download, install and use it at no cost.
Which platforms does BEE Belief Extraction Engine support?
BEE Belief Extraction Engine is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created BEE Belief Extraction Engine?
It is built and maintained by Kartik Vashisth (@vashkartik); the current version is v0.1.1.
More Skills