ClawHub
Back to skill

Security audit

Sombra — Reader Mode for AI

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed Sombra MCP skill for saving and reusing research context, with manageable persistence and token-handling risks but no evidence of malicious behavior.

Install only if you trust Sombra and the sombra-mcp package. Treat the Sombra token as a secret, prefer OAuth or safer secret storage if available, review agent write/delete actions for important collections, and avoid saving credentials, regulated data, or sensitive private material unless you are comfortable storing and reusing it through Sombra.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

Medium
Confidence
85% confidence
Finding
The workflow examples are written as generic natural-language tasking ('Research...', 'Save...', 'write a distilled context...') without clear boundaries indicating when the skill should or should not be invoked. In agentic environments, this can cause overbroad activation during ordinary research requests, leading the agent to persist fetched content or send data to the external Sombra MCP service unexpectedly.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The skill promotes persistent saving of web pages, notes, and distilled context, but the introductory description does not clearly warn that this data persists across sessions and is handled by an external hosted MCP service. This increases the risk that users or downstream agents will store sensitive internal data, credentials, or proprietary research without informed consent about retention and external transmission.

Static analysis

No suspicious patterns detected.