Back to skill

Security audit

Crypto Research

Security checks for vulnerabilities and agentic risk

Overview

The skill provides crypto research through CellCog, with expected privacy, software installation, and background notification considerations.

CellCog receives the prompts you submit, so minimize identifying portfolio details and never include wallet secrets. The advisory privacy and unpinned-installation concerns warrant these precautions, but do not establish hidden or excessive behavior. Review dependencies before installation, understand API credit costs and background notifications, and independently verify research before acting.

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)

Task prompts containing sensitive portfolio information are sent to CellCog without a privacy warning.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The skill sends task prompts to an external CellCog service and explicitly encourages portfolio reviews containing holdings, valuations, and financial preferences, without explaining the privacy boundary or requiring data minimization. This is consistent with the advertised research workflow rather than evidence of malicious exfiltration, but it can disclose sensitive financial information when an agent forwards user details without informed authorization.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding

Although this command is a skill installer rather than an MCP server reference, the underlying supply-chain concern is valid: npx skills can download and execute an unpinned package, and the referenced skill source is also not pinned to an immutable revision. A compromised package release or upstream skill update could introduce code or instructions that run with the agent user's privileges; nothing in the supplied file establishes that such compromise has occurred.

Content

No source excerpt is available for this finding.