Clawguard Self Improving Safety

Self-improving safety module that proactively learns and enforces safety constraints. Works as a standalone firewall or integrates with ClawGuard Detect for automated learning.

49 stars 4 forksLearning & Education

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add SafeAgent-Beihang/clawguard --skill clawguard-self-improving-safety
Manual install
git clone --depth 1 https://github.com/SafeAgent-Beihang/clawguard.git /tmp/clawguard
mkdir -p ~/.claude/skills
cp -r /tmp/clawguard/self-improving-safety ~/.claude/skills/clawguard-self-improving-safety

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# ClawGuard Self-Improving Safety

A proactive safety enforcement module that learns from detected threats. It consumes alerts from **ClawGuard Detect**, analyzes patterns, and automatically updates its blocklist to prevent recurrence of dangerous activities.

## When to Use

Activate ClawGuard Self-Improving Safety when:
- You want the agent to learn from its mistakes or detected attacks.
- You need a dynamic defense mechanism that evolves with new threats.
- Enforcing strict safety protocols based on historical incident data.

## How it Works

1.  **Listen**: Monitors output from `clawguard-threat-detect`.
2.  **Learn**: When a threat is detected, it extracts a blocking rule (e.g., specific command pattern, sensitive file path).
3.  **Enforce**: Updates the `BLOCKLIST.json` which is checked before executing any agent action.

## Integration & Examples

Can be used independently or integrated with detection systems like `clawguard-threat-detect`.

### Standalone Examples

**1. Block dangerous commands:**
```bash
node cli.js add-rule command "rm -rf /"
node cli.js add-rule command "format c:"
```

**2. Protect sensitive files:**
```bash
node cli.js add-rule file "/etc/shadow"
node cli.js add-rule file "id_rsa"
```

**3. Check input against rules:**
```bash
node cli.js check "rm -rf /user/data"
# Output: BLOCKED
```

### Integrated Workflow

Pipeline detection output to safety module for automatic learning:
Read the full skill on GitHub

Details

Skill name
clawguard-self-improving-safety
Path
self-improving-safety/SKILL.md
Repo updated
29 March 2026

FAQ

What does the Clawguard Self Improving Safety skill do?

Self-improving safety module that proactively learns and enforces safety constraints. Works as a standalone firewall or integrates with ClawGuard Detect for automated learning.

How do I install the Clawguard Self Improving Safety skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add SafeAgent-Beihang/clawguard --skill clawguard-self-improving-safety. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Clawguard Self Improving Safety popular?

It lives in SafeAgent-Beihang/clawguard, which has 49 GitHub stars and 4 forks.

Related skills

Clawguard Auditor

49

SafeAgent-Beihang/clawguard

ClawGuard v3 Auditor - 企业级 Skill 安全审计器,支持意图偏离检测、SAST、供应链安全、ML 异常检测。当用户要求审计、检测、安装前检查一个 Skill 的安全性时触发。

AI, LLMs & Agents

Clawguard Guardian

49

SafeAgent-Beihang/clawguard

ClawGuard Guardian v3 - Runtime guardian with behavior monitoring, interception, session freeze/replay, and emergency response

DevOps & Cloud

Clawguard Security Checker

49

SafeAgent-Beihang/clawguard

ClawGuard Security Checker v3 - Advanced configuration analysis, runtime integrity verification, permission modeling, and one-click hardening recommendations

Security

Smart Donkey Learn

43

cloudyview/smart-donkey

Trigger Smart Donkey's learning phase: distill insights from the current session into smart-donkey-brain.md.

Learning & Education

Shatter

113

Hmbown/Wizards-of-the-Ghosts

Shatter is proactive, controlled destruction for learning. You are the attacker against your own system. The goal is a brittleness report with remediation priorities.

Learning & Education11

Feedback Coach

48

harvard-lil/lawskills-hub

Checks an instructor's draft feedback on a student response against the assignment's grading rubric and rewrites it to be constructive, specific, and accessible to a neurodiverse audience. Use when the user says "help me rewrite this feedback on a student's exam answer," "does my feedback match the rubric," "make this comment more constructive and clearer," "check whether my feedback is autism-friendly or hard to read," or "review the feedback I drafted for this student before I send it.

Learning & Education

Browse the most popular Claude skills, all learning & education skills, or write your own with the skills generator.