Back to skill

Security audit

YouTube Video

Security checks for vulnerabilities and agentic risk

Overview

This skill coherently delegates video creation to CellCog, with ordinary third-party privacy, cost, and dependency-installation considerations.

Prompts and content included in them go to CellCog, so share only material you are authorized to send. Set a generation budget and review the separately installed CellCog dependency and its provenance before supplying your API key. This assessment covers the supplied instructions, not the external SDK or service implementation.

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 workflow sends task prompts to CellCog without warning users that their content leaves the local agent environment.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The usage instructions direct agents to submit the user's task prompt through CellCog's create_chat API. Although CellCog is identified as the provider, the file does not disclose that prompts and any user content included in them are transmitted to a third-party service; the later warnings address cost and output quality, not privacy.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The installation command invokes an unpinned npm package through npx and references skill content without an immutable revision, allowing upstream changes to affect what is executed or installed. A compromised upstream release could execute code with the invoking user's permissions; however, the command is conditional setup guidance rather than automatic execution. The supply-chain concern is valid, although the finding incorrectly describes this as an MCP server reference.

Content

No source excerpt is available for this finding.