Security audit

Manus on OpenClaw

Security checks across malware telemetry and agentic risk

Overview

This Manus bridge mostly matches its stated purpose, but one slide-conversion path can fetch and write files outside the documented safety limits, so it should be reviewed before installation.

Install only if you trust the Manus endpoint and publisher, use a dedicated Manus API key, protect the local config file, and avoid running the slides JSON-to-PPTX converter on untrusted output until it enforces HTTPS Manus-host allowlisting, redirect revalidation, size limits, and safe filenames.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Vague Triggers

Medium

Confidence: 83% confidence
Finding: The guidance recommends a simple direct-message trigger that strips prefixes like 'Manus,' or 'manus:' and forwards the remainder, without defining authentication, confirmation, sender scoping, or explicit exclusions. In chat-driven automations, broad trigger rules can cause accidental or unauthorized task launches, potentially sending sensitive user text to an external API and consuming paid resources.

Missing User Warnings

Medium

Confidence: 79% confidence
Finding: The script automatically downloads remote files to disk without prompting the user, enforcing size limits, or validating content beyond hostname checks. In this skill context, task outputs are expected, but a compromised upstream service or malicious task content could cause storage exhaustion or place risky files on disk for later user handling.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal