ClawHub
Back to skill

Security audit

Office Collaboration Radar (English)

Security checks across malware telemetry and agentic risk

Overview

The skill is a local collaboration-summary tool with manageable auto-invocation and file-overwrite caveats, and I found no evidence of hidden exfiltration or system-changing behavior.

Install only if you are comfortable letting the agent summarize supplied collaboration records that may contain business-sensitive information. Prefer explicit invocation for sensitive chats or meeting notes, and use explicit --out paths so draft card files are not overwritten accidentally.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

High
Confidence
90% confidence
Finding
The skill metadata says it is read-only by default, but the CLI writes output and will overwrite the input card file when --out is omitted. In an agent setting, that mismatch can cause unintended modification or loss of user data, especially if higher-level orchestration assumes the tool is non-mutating and invokes it on trusted files.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The skill enables implicit invocation without any visible trigger constraints or narrowing conditions, so the agent may auto-select this skill in contexts broader than intended. Because the skill processes collaboration records and produces structured summaries, unintended activation could expose sensitive business context, cause data to be processed under the wrong workflow, or produce misleading summaries from unrelated inputs.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The enforce command writes to disk without any confirmation, and by default targets the same path provided in --card. That creates a silent overwrite risk: a caller or user can lose the original draft card or have data altered unexpectedly, which is especially problematic for security-sensitive pipelines that expect inspection before persistence.

Missing User Warnings

Low
Confidence
77% confidence
Finding
The aggregate command creates or overwrites an output file without a warning, defaulting to aggregated.json when --out is not specified. This is less severe than in-place overwrite of an input file, but it can still clobber existing artifacts or create unexpected files in an automated workspace.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.