Clawkeeper vs ClawSec
How Clawkeeper compares to ClawSec by Prompt Security, an OpenClaw skill suite for advisory monitoring and skill reputation.
ClawSec runs inside the agent to monitor skill advisories. Clawkeeper runs outside the agent to scan the entire deployment stack — host OS, network, containers, config, and credentials. They're complementary, not competing.
At a Glance
| Clawkeeper | ClawSec | |
|---|---|---|
| Type | External scanner + dashboard | OpenClaw skill (runs inside agent) |
| Checks | 55 automated across 5 phases | ~12 (advisory + reputation + audit) |
| Scope | Full-stack: host + network + container + config | OpenClaw-only: skills + advisories |
| Platform | macOS + Linux (bare metal, Docker, K8s) | Any OS with Node.js |
| Remediation | Interactive auto-fix for 20+ issues | Advisory-gated skill removal only |
| Dashboard | Pro web dashboard with fleet monitoring | None (CLI output only) |
| Deployment | Independent bash script (zero deps) | Installed as OpenClaw skill |
| Pricing | Free (1 workstation) / Pro $19/mo / Team $29/seat/mo (min 3) / Enterprise | Free (open source) |
Feature-by-Feature Comparison
| Feature | Clawkeeper | ClawSec |
|---|---|---|
| Host OS hardening (firewall, disk encryption, users) | ||
| Network security (mDNS, ports, SSH, screen sharing) | ||
| Container security (13 Docker sub-checks) | ||
| Gateway config audit (bind, auth, UI, mDNS) | ||
| Advanced gateway (elevated tools, browser, proxies) | ||
| Sandbox & execution policy | ||
| DM scope & policy | ||
| Credential scanning (config, history, memory, sessions) | ||
| Credential store permissions | ||
| SOUL.md integrity (injection, steganography, base64) | ||
| Skills static analysis (install cmds, exfiltration) | ||
| .env file security | ||
| Session prompt injection detection | ||
| Session rogue command scanning | ||
| MEMORY.md prompt injection detection | ||
| Skills prompt injection detection | ||
| Log file content scanning | ||
| CVE / version vulnerability check | ||
| Cryptographic advisory feed | ||
| Skill reputation scoring (7-check system) | ||
| Guarded skill installation hook | ||
| Real-time advisory monitoring (heartbeat) | ||
| Advisory suppression config | ||
| A-F letter grade + scoring | ||
| Interactive auto-fix (20+ remediations) | ||
| Fleet monitoring dashboard | ||
| AI-powered security insights | ||
| Email & webhook alerting | ||
| Score trend tracking | ||
| Skill Marketplace (13K+ community skills) | ||
| Enterprise plan with SSO & SLA |
Where ClawSec Excels
Cryptographic advisory feed — ClawSec maintains a signed JSON feed with Ed25519 signatures and SHA-256 checksums, automatically polling NVD for new OpenClaw CVEs via GitHub Actions.
Skill reputation scoring — A 7-check system (existence, age, staleness, author reputation, downloads, VirusTotal, version validation) with a 0-100 score and configurable threshold.
Guarded installation — Intercepts clawhub install to block unsafe skills before they're installed. Clawkeeper scans post-installation.
Real-time monitoring — Hook-based heartbeat runs every 5 minutes inside the agent, checking for new advisories matching installed skills.
Where Clawkeeper Excels
Full-stack coverage — 55 checks across host OS, network, containers, OpenClaw config, credentials, skills, CVEs, and SOUL.md. ClawSec only covers skills and advisories.
External trust boundary — Clawkeeper runs outside the agent as an independent bash script. If the agent is compromised, the scanner still works. ClawSec runs inside the agent.
Interactive remediation — Auto-fix for 20+ issues with user confirmation: firewalls, permissions, SSH hardening, skill quarantine, and more. ClawSec only removes skills matching advisories.
Fleet dashboard — Centralized web dashboard with historical trends, A-F grades, alert rules, and team management. ClawSec has no monitoring UI.
Enterprise path — Unlimited hosts, SSO/SAML, dedicated support, and SLA. ClawSec has no enterprise offering.
ClawSec is an OpenClaw skill — it has no support for Claude Code whatsoever. Clawkeeper is the only platform providing real-time security monitoring for Claude Code workstations.
| Feature | Clawkeeper | ClawSec |
|---|---|---|
| Claude Code Runtime Shield (55+ patterns, PreToolUse blocking) | ||
| Claude Code workstation fleet management | ||
| Claude Code team policies (blocked tools, paths, commands, skills) | ||
| MCP skill monitoring and policy enforcement | ||
| GitHub integration (auto-PR hooks config) | ||
| JAMF / MDM deployment support | ||
| Identity provider sync (Entra, Google, Okta) |
ClawSec operates entirely within the OpenClaw agent runtime. It cannot monitor, manage, or enforce policies on Claude Code sessions.
Recommendation
Use both. ClawSec and Clawkeeper are complementary — they cover different layers of the security stack with essentially zero overlap. ClawSec handles real-time advisory monitoring and supply-chain gating inside the agent. Clawkeeper handles full-stack security posture scanning from the outside. For Claude Code security, Clawkeeper is the only option.