← Back to Skills Marketplace
Openclaw Sec
by
PaoloRollo
· GitHub ↗
· v0.2.6
5239
Downloads
10
Stars
16
Active Installs
8
Versions
Install in OpenClaw
/install openclaw-sec
Description
AI Agent Security Suite - Real-time protection against prompt injection, command injection, SSRF, path traversal, secrets exposure, and content policy violations
Usage Guidance
Install only if you want OpenClaw-wide security hooks that may inspect and log prompts and tool-call parameters. Review the postinstall hook behavior, database path and retention, owner bypass settings, notification endpoints, and whether `tool_result_persist` truly blocks before execution in your OpenClaw version.
Capability Analysis
Type: OpenClaw Skill
Name: openclaw-sec
Version: 0.2.6
The OpenClaw Security Suite is a security tool designed to detect and prevent various attacks like prompt injection, command injection, SSRF, path traversal, and secret exposure. All analyzed files, including the `SKILL.md` documentation, TypeScript source code, and shell scripts (`hooks/install-hooks.sh`), consistently align with this stated purpose. The tool's functionalities, such as scanning for malicious patterns, logging security events to a local SQLite database (`.openclaw-sec.db`), and sending notifications to configurable endpoints, are legitimate for a security product. The 'malicious' examples found in the code and documentation are test cases that the tool is designed to *detect and block*, not execute. There is no evidence of intentional harmful behavior, data exfiltration to unauthorized external endpoints, or persistence mechanisms beyond standard OpenClaw hook installation.
Capability Assessment
Purpose & Capability
The scanning, blocking, reputation, database logging, and optional notifications generally fit the stated security-suite purpose; no artifact-backed exfiltration or destructive behavior was found.
Instruction Scope
The tool-call hook claims pre-execution blocking while registering a `tool_result_persist` plugin, creating real ambiguity about whether tool actions are blocked before or after execution.
Install Mechanism
`package.json` runs a postinstall script that builds the package and invokes `hooks/install-hooks.sh`, which copies hooks into `~/.openclaw/hooks`, creates a symlink, and attempts to enable them via the OpenClaw CLI.
Credentials
Default-enabled hooks can inspect user prompts and security-relevant tool parameters, then store normalized input, user IDs, session IDs, findings, and metadata in SQLite for analytics; this is purpose-aligned but sensitive and broad.
Persistence & Privilege
The hooks persist beyond the install command and alter agent behavior globally under the OpenClaw hooks directory; documentation includes disable steps, but the install-time side effect is high-impact.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install openclaw-sec - After installation, invoke the skill by name or use
/openclaw-sec - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.2.6
openclaw-sec 0.2.6
- Added new detection modules: code execution, exfiltration, injection validator, and serialization detection.
- Introduced corresponding runtime validation patterns for code execution, exfiltration, SQL injection, template injection, general injection, and serialization attacks.
- Significantly expanded test coverage for all new modules and patterns.
- Updated core security engine and type definitions to support new modules.
- Enhanced the modular structure to accommodate more security pattern categories.
v0.2.5
- Added pnpm-lock.yaml for improved dependency management and reproducible builds.
- Updated multiple test files for integration and CLI functionality.
- Modified Jest configuration for testing enhancements.
- Made changes to the action engine and notification system test modules.
v0.2.4
openclaw-sec 0.2.4 Changelog
- Updated capability metadata: added "168 patterns across 16 categories" to detection coverage.
- SKILL.md version updated from 1.0.0 to 1.0.2, reflecting underlying metadata improvements.
- No user-facing changes to commands or functionality. Documentation aligned with current detection scope.
v0.2.3
openclaw-sec v0.2.3
- Added 7 new pattern detectors for advanced prompt injection types, including chain-of-thought hijacking, direct extraction, encoding obfuscation, extraction attacks, policy puppetry, and social engineering.
- Integrated these patterns into the core prompt injection module.
- Added automated continuous integration (CI) workflow for better testing.
- Improved and expanded test coverage for action and security engines.
- Removed obsolete benchmark result files to streamline the codebase.
- Updated documentation and dependencies.
v0.2.2
- Added `.openclaw-sec.example.yaml` as the new example config file and removed the old `.openclaw-security.example.yaml`.
- Updated and improved documentation in `README.md`, hooks, and contributing guides.
- Refined configuration examples and inline docs to align with new file naming.
- Various updates across codebase and tests to reflect config changes and improve clarity.
- No breaking changes to commands or core features.
v0.2.1
- Updated documentation in README.md and package.json.
- Removed the PROJECT_SUMMARY.md file for simplification.
- No changes to core functionality or detection modules.
v0.2.0
Summary: Major expansion of test coverage and security hooks; legacy hook refactor.
- Added extensive integration and performance test suites.
- Introduced new security input/tool validator hooks and documentation.
- Split legacy hooks into a dedicated subfolder with improved structure.
- Removed legacy hook files in favor of modular handlers.
- Updated and clarified documentation and contribution guidelines.
v0.1.0
Initial release of OpenClaw Security Suite – comprehensive, real-time security for AI agents.
- 6 parallel detection modules cover prompt injection, command injection, SSRF, path traversal, secrets exposure, and content policy violations.
- Real-time validation (<50ms) with async database/logging.
- Smart severity scoring and automated action enforcement (block, warn, log).
- Provides detailed CLI/tools for validation, scanning, monitoring, and analytics.
- Includes user reputation tracking, security stats, and event monitoring.
Metadata
Frequently Asked Questions
What is Openclaw Sec?
AI Agent Security Suite - Real-time protection against prompt injection, command injection, SSRF, path traversal, secrets exposure, and content policy violations. It is an AI Agent Skill for Claude Code / OpenClaw, with 5239 downloads so far.
How do I install Openclaw Sec?
Run "/install openclaw-sec" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Openclaw Sec free?
Yes, Openclaw Sec is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Openclaw Sec support?
Openclaw Sec is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Openclaw Sec?
It is built and maintained by PaoloRollo (@paolorollo); the current version is v0.2.6.
More Skills