Back to skill

Security audit

Dashboard

Security checks for vulnerabilities and agentic risk

Overview

This skill uses CellCog to build dashboards and apps, with expected precautions around sharing data and installing dependencies.

Prompts and uploaded files go to CellCog; use sample or minimized data and share sensitive records only when authorized. Check the service's data policies, protect CELLCOG_API_KEY, and review the separate CellCog skill and SDK before setup. Installation commands are unpinned, so verify their sources and versions. This review covers the supplied instruction file, not downloaded dependencies or remote service behavior.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

The data-upload workflow lacks a warning that user data is sent to the external CellCog service.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The workflow encourages providing inline datasets and uploaded files without explicitly explaining that their contents are transmitted to an external service. Examples involving employee salaries, financial records, and logs make accidental disclosure consequential, although the CellCog branding provides some context and there is no evidence of covert or automatic exfiltration.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

Although this is a CLI installer rather than an MCP server, the underlying supply-chain concern is valid: npx skills can download and execute an unpinned package version with the invoking user's permissions. A compromised release could access local files or credentials; the fact that this is an optional setup step limits exposure but does not remove the risk.

Content

No source excerpt is available for this finding.