ClawHub
Back to skill

Security audit

skill-compass

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed tool for auditing and repairing skill descriptions, with expected file access and backups, but users should be careful before enabling automatic rewrites or rollback.

Install only if you want an agent to inspect and potentially modify your installed skill descriptions. Run audit mode first, pass an explicit --skills-dir, review generated rewrite prompts before applying them, and keep backups before using --fix, --init, or --rollback.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (6)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill instructs the agent to run audit and fix scripts and to mark onboarding complete, which implies file read/write behavior, but it does not declare any permissions or safety boundaries for those capabilities. This can lead to unexpected modification of skill files or state during troubleshooting, especially because the skill also promotes auto-remediation flows.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The example trigger phrase “检查一下我的 skills” is a broad, natural-language request that could cause the skill to activate in routine conversations beyond explicit troubleshooting intent. Because this skill is designed to audit and potentially modify skill descriptions, over-broad triggering increases the chance of unnecessary autonomous actions or unintended rewrites in agent environments.

Vague Triggers

Medium
Confidence
85% confidence
Finding
The activation text is very broad, covering generic phrases like 'skill not working,' 'agent behavior seems to ignore available skills,' and proactive auditing after installs. Because many unrelated support conversations could match these phrases, the skill may trigger too often, creating routing interference and causing the agent to follow invasive audit/fix workflows in contexts where they were not intended.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
This code can modify SKILL.md files automatically when auto-fix is enabled, and it does so without an explicit confirmation step immediately before writing. Even though the feature appears intended for convenience and a backup is created in some CLI flows, unattended or mistaken invocation can still alter user content and create integrity issues.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The rollback routine overwrites destination SKILL.md files directly from backup contents without prompting the user to confirm the restore set. If the wrong skills_dir or backup is selected, this can silently replace current files and cause data loss or reintroduce unsafe content from an older backup.

Session Persistence

Medium
Category
Rogue Agent
Content
**Bad:** `"Helps with documents."` — no trigger, just capability
**Good:** `"Use when the user asks to extract form fields, fill, redact, or parse tables from a PDF file."`

**Fix:** Rewrite using the Description Design Pattern below.

### 2. YAML Frontmatter Format Error
Confidence
62% confidence
Finding
write using the Description Design Pattern below. ### 2. YAML Frontmatter Format Error Common breakages: - Colon in description breaks YAML parsing (GitHub #29981) - Missing `name:` or `description:

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.