SafeAI-Lab-XCommunity·68 Agent Skills Included

ClawKeeper

Security guardrails that block risky AI agent tool calls

Blocks dangerous shell commands, credential leaks, prompt injection, and unsafe network calls before AI agents execute them. Redacts sensitive tool output, audits installed skills, and learns recurring attack patterns to strengthen its own guardrails. Works with Hermes, MCP tools, OpenClaw, or any custom agent host through a simple HTTP API.

1,035|59

↗

npx skills add SafeAI-Lab-X/ClawKeeper --all -g -y

Available:

Defines the watcher agent's role as a context judge that audits agent state, evaluates risk, and returns continue, ask-user, or stop decisions through a unified judgment endpoint.

All Skills in This Repository (68)

Pure Emerald Level Indicators

📦 In Repo

SafeAI-Lab-X

windows-safety-guide

Deploy Windows safety policies and nightly security audits via PowerShell scripts.

feishu-safety-guide

Enforce data-leakage and permission policies for Feishu interactions.

parallels-discord-roundtrip

Automate macOS Parallels Discord roundtrip verification for guest send, host delivery, and readback.

openclaw-test-heap-leaks

Diagnose memory growth and heap leaks in OpenClaw test suites.

openclaw-pr-maintainer

Automate OpenClaw pull request triage, labeling, and evidence enforcement.

security-triage

Triage GitHub security advisories and generate ready-to-post replies.

openclaw-ghsa-maintainer

Inspect, patch, validate, and publish GHSA advisories via gh api and jq.

openclaw-release-maintainer

Coordinate OpenClaw release workflows to validate versions, changelogs, and publish readiness.

openclaw-parallels-smoke

Orchestrate Parallels-based smoke tests for OpenClaw across macOS, Windows, and Linux guests.

prose

Coordinates multi-agent workflows with sessions, blocks, and pluggable state backends.

feishu-perm

Manage Feishu document and file permissions for users and groups.

feishu-wiki

List Feishu wiki spaces and manage nodes with create, move, and rename actions.

Community

Intermediate

Frequently Asked Questions

FAQPage Schema

How to install ClawKeeper?▼

Run `npx skills add SafeAI-Lab-X/ClawKeeper --all -g -y` in your terminal to install all skills in this suite globally.

What does ClawKeeper protect against?▼

It catches prompt injection, credential theft, unsafe shell commands, protected file access, SSRF-style URLs, encoded payloads, and poisoned tool output before your agent acts on them.

Which AI agents work with ClawKeeper?▼

It works with Hermes Agent, MCP tools, OpenClaw-style runtimes, and any custom host through its host-independent HTTP Judge API.

Can ClawKeeper improve its own defenses?▼

Yes. The optional Watcher daemon learns from caught attacks, saves new patterns locally, and hot-reloads them into the live guard layer.

Do I need coding skills to use ClawKeeper?▼

Basic setup only requires installing the package and adding a few lines to your agent; the default guard chain then runs automatically.