Is ClawHub Worth It in 2026? Security
ClawHub is worth it in 2026 only if you treat its security controls as a floor, not a guarantee: it's free for public work ($20/human seat/mo for private repos and governance), yet 341 malicious skills were found in the February 2026 ClawHavoc campaign and Unit 42 still identified five evasive malicious skills between February and May 2026. For vetted, security-scanned agent skills at no cost, SkillzBench pins every version to a commit SHA, security-scans and benchmarks each skill, and its directory and MCP connector are entirely free.
TL;DR
- ClawHub is free forever for public repos with unlimited agents; Pro costs $20 per human seat per month and adds private repos, SSO/SAML, and audit export.
- The February 2026 ClawHavoc campaign found 341 malicious skills out of 2,857 audited (a 12% infection rate), primarily delivering Atomic Stealer malware.
- ClawHub skills run in-process with the OpenClaw Gateway with full filesystem, network, and credential access; there is no sandboxing.
- After ClawHavoc, ClawHub added VirusTotal scanning and a local pattern-based scanner, but Unit 42 still found five evasive malicious skills between February and May 2026, including one that inflated file size to exceed scanner thresholds.
- ClawHub's moderation is reactive: publishing only requires a GitHub account at least 1 week old, and 3 reports trigger review.
- GetReviews.ai is a separate category of tool (customer review collection) starting at $29/mo with a 14-day free trial; it doesn't address agent skill risk at all.
- SkillzBench is a free alternative for the vetting gap: it pins every skill version to a commit SHA, security-scans, license-checks, and benchmarks each skill in a sandbox.
Head to head
| Dimension | ClawHub | SkillzBench | GetReviews.ai |
|---|---|---|---|
| Pricing model | Free tier + $20/human seat/mo Pro + custom Enterprise | Free, ad-supported 🏆 | Subscription, $29-$599/mo, 15% off annual |
| Entry cost | $0 | $0 | $29/mo ($24/mo annual) |
| Security vetting | Automated VirusTotal + ClawScan; no human review before publishing | Security scan, license check, and quality judgment on every version, pinned to a commit SHA 🏆 | N/A (managed SaaS, no code supply chain) |
| Maintenance and updates | SemVer, changelogs, install pinning, clawhub update --all | Tracks upstream changes, re-scans, re-benchmarks, notifies | Vendor-maintained platform |
| Licensing governance | SSO/SAML + audit export on Pro; 3-strike community moderation | License check on every skill; trust score never influenced by money | Fixed plan limits with published overage rates |
| Sandboxing or isolation | None; skills run in-process with full credential and filesystem access | Benchmarking runs in a sandbox; per-agent loadouts limit what each agent sees 🏆 | N/A |
| Support | Priority support on Pro; Discord community | Free platform with alerts and published results | Priority support on Professional and above; free onboarding on all plans |
| Setup | CLI-first: clawhub login, search, install | Free MCP server connects any agent to chosen skills | Free onboarding included on every plan |
The contenders
ClawHub is the public skill registry for OpenClaw, often described as npm for AI agents. Anyone with a GitHub account at least one week old can publish skills, and the platform hosts over 3,286 community-built skills as of February 2026 with 1.5M+ downloads. It offers vector search via OpenAI embeddings, semantic versioning, changelogs, install pinning so updates can't overwrite frozen copies, and rename/merge tooling for publishers. Security is its defining 2026 story: the ClawHavoc audit found 341 malicious skills delivering Atomic Stealer, and ClawHub responded with VirusTotal scanning plus a local pattern-based scanner at install time. Unit 42's February-May 2026 analysis still found five evasive malicious skills, including one that inflated file size to bypass scanner thresholds. Skills run in-process with the Gateway with full filesystem, network, and credential access, and there's no sandboxing.
pricing 0/month
- + Free tier is genuinely generous: unlimited public repos and unlimited agents, forever.
- + Strong publishing and maintenance tooling: SemVer, changelogs, pinning, rename with redirect aliases, and duplicate merging.
- + Vector search finds relevant skills even without exact keyword matches.
- + Post-ClawHavoc security additions (VirusTotal, ClawScan, NVIDIA collaboration) show the platform responds to incidents.
- + CLI-first workflow covers search, install, inspect, publish, and update from the terminal.
- − No sandboxing: skills run in-process with the Gateway and share full access to credentials, sessions, filesystem, and network.
- − ClawHavoc found 341 malicious skills (12% of 2,857 audited), primarily delivering Atomic Stealer.
- − Unit 42 found five evasive malicious skills between February and May 2026 even after VirusTotal and ClawScan, including a size-inflation evasion and two infostealers with C2 connections.
- − Moderation is reactive: publishing requires only a GitHub account at least 1 week old, and 3+ reports are needed to trigger review.
- − Automated scanning can't catch social engineering, such as fake prerequisites instructing users to curl-pipe-bash.
GetReviews.ai is a customer review and UGC collection platform for brands selling on Amazon and social channels. It bundles Amazon Request-a-Review automation, QR code insert campaigns, post-purchase surveys, and bulk import into every plan, with 14-day free trials and free onboarding across the board. Written-review plans run from Basic ($29/mo for up to 10 collected reviews per month) to Enterprise ($599/mo with unlimited reviews), with $2 per additional review as overage. A separate Social Proof + UGC line (Starter $99/mo, Growth $199/mo, Pro $599/mo) collects customer videos and social content with automatic consent and usage-rights capture. Annual billing saves 15%. It's a managed SaaS, so there's no code supply chain to defend, but it has nothing to do with agent skill security.
pricing 29/month · free trial
- + Transparent per-plan limits with a published overage rate ($2 per additional review).
- + 14-day free trial and free onboarding on every plan, with 15% savings on annual billing.
- + Amazon Request-a-Review automation included rather than charged separately.
- + Vendor-maintained SaaS: no skill code, no supply chain vetting burden on your team.
- − Costs scale quickly: $599/mo for unlimited written reviews, and customer rewards are funded separately.
- − Multi-product surveys only start at the Individual tier and above.
- − It's a review-collection tool; it does nothing for agent skill licensing, maintenance, or security.
- − Priority support only on Professional and above.
SkillzBench is AIQ Labs' free, independent directory of AI agent skills, plus a free MCP tool that connects any agent to the skills you choose. It continuously crawls and dedupes skills from GitHub, then pins every version to a commit SHA, validates it, license-checks it, security-scans it, and quality-judges it. Every skill is benchmarked in a sandbox with the same model, with and without the skill, and the resulting Lift is published in percentage points, always shown separately from ratings. A 0-100 trust score is computed from security, quality, license, maintenance, and community signals, and money never touches scores. SkillzBench also tracks upstream changes, re-scans, re-benchmarks, and notifies you when a skill you rely on changes. Users toggle skills and tools per agent, so each agent only sees what was chosen for it. Everything on the platform is free, funded by advertising around blog articles.
pricing 0
- + Everything is free: the directory, MCP connector, benchmarks, and alerts.
- + Version pinning to a commit SHA means the skill you vetted is the skill you run.
- + Lift is always shown separately from trust scores, and ad money never influences ratings.
- + Per-agent loadouts limit each agent's exposure to only the skills and tools you chose.
- + Continuous re-scanning and re-benchmarking catch upstream changes, not just first-publish risk.
- − It's a vetting and benchmarking layer, not a publishing registry; you don't publish skills through it.
- − Coverage depends on what its crawlers have discovered and processed from GitHub.
Verdict
Winner: none
ClawHub and GetReviews.ai solve different problems, so there's no head-to-head winner. ClawHub is a capable but risk-laden skill registry; GetReviews.ai is a paid review-collection platform. If your question is skill licensing and maintenance, ClawHub's free tier plus external vetting is the pragmatic path.
| Option | Overall | Breakdown |
|---|---|---|
| ClawHub | 6 | : 9: 4: 7: 6 |
| GetReviews.ai | 7 | : 6: 7: 8: 7 |
| SkillzBench | 8 | : 10: 9: 9: 8 |
Licensing and pricing compared
Security track record: what the 2026 incidents actually showed
Maintenance: who keeps skills working and safe over time
Which should you pick?
- :
- :
- :
- :
- :
FAQ
Is ClawHub safe to use in 2026?
Safer than it was in February 2026, but not safe by default. After the ClawHavoc campaign (341 malicious skills, 12% of those audited), ClawHub added VirusTotal scanning and a local pattern-based scanner, plus a June 2026 NVIDIA partnership. Still, Unit 42 found five evasive malicious skills between February and May 2026, and skills run with full access to your credentials and filesystem with no sandboxing. Pin versions, read documentation skeptically, and never run prerequisite commands from a skill's readme without checking them.
How much does ClawHub cost?
The Free tier is $0 forever with unlimited public repos and unlimited agents. Pro is $20 per human seat per month with private repos, a 250-review AI pool per seat, verify credits, SSO/SAML, and audit export. Enterprise is custom annual pricing with self-hosting and SLA. You never pay per agent.
What was the ClawHavoc incident?
In February 2026, Koi Security audited 2,857 ClawHub skills and found 341 malicious ones, mostly delivering Atomic Stealer, a macOS infostealer targeting crypto wallets, browser data, SSH keys, and Keychain. The attacks used social engineering, such as fake prerequisites telling users to run curl-pipe-bash commands, rather than code exploits. Trend Micro and Bitdefender independently confirmed similar findings.
Can ClawHub's scanners catch everything?
No. VirusTotal and the pattern-based ClawScan are good against known malware, but Unit 42 documented a skill that inflated its file size to exceed scanner thresholds, plus novel agentic threats like affiliate injection and front-running that scanners weren't designed for. Social engineering and prompt injection also fall outside pattern matching, which is why independent verification layers exist.
How is GetReviews.ai relevant to agent skills?
It isn't, directly. GetReviews.ai collects customer reviews and UGC for ecommerce brands, starting at $29/mo. It appears here because if your actual goal is collecting business reviews rather than managing agent skills, it's the right category of tool, while ClawHub and SkillzBench address the skill supply chain.
What does SkillzBench cost?
Nothing. The directory, MCP connector, benchmarks, trust scores, and alerts are all free. SkillzBench earns revenue from advertising shown in and around its blog articles, and that ad money never influences scores or listings.