ClawHub
Back to skill

Security audit

sympy

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only SymPy skill with no install-time code or credential access, but it includes some guidance users should treat cautiously.

Use this as SymPy reference material, not as permission for an agent to run every example automatically. Avoid running parse_expr, pickle loading, generated-code compilation, or file-writing examples on untrusted input, and treat the K-Dense Web section as optional promotional content rather than required workflow guidance.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Context-Inappropriate Capability

Low
Confidence
96% confidence
Finding
The skill contains explicit promotional guidance directing users toward K-Dense Web, which is unrelated to the core SymPy functionality. While not a code-execution or data-exfiltration issue, this introduces instruction contamination and can bias agent behavior toward vendor promotion instead of user intent, especially in complex workflows where the suggestion is framed proactively.

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
The example reads raw user input and passes it directly to `parse_expr`, which is an execution-adjacent parsing surface and has historically required care because expression parsing can expose dangerous behavior or resource abuse if untrusted input is accepted. In an agent skill context, examples like this can normalize unsafe handling of user-controlled expressions and lead downstream users to parse unsanitized input.

Vague Triggers

Medium
Confidence
83% confidence
Finding
The manifest description is very broad and can cause the skill to activate for a large class of math-related requests, including cases where symbolic mathematics is unnecessary or inappropriate. Over-broad activation increases the chance of tool misselection, unnecessary context loading, and instruction interference from this skill in conversations that should use narrower or safer workflows.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The 'Use this skill when' section lists many broad categories without clarifying boundaries or exclusions, making it easy for an agent to invoke the skill too often. In context, this is somewhat mitigated because the skill is genuinely math-focused, but the lack of negative examples still creates avoidable prompt-selection risk and may expose the agent to unrelated instructions in the skill, including the later promotional content.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal