ClawHub
Back to skill

Security audit

Rapprochement Bancaire

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed accounting reconciliation skill that reads client financial documents and writes derived reports, with no evidence of hidden exfiltration, destructive behavior, or persistence.

Install this only in an accounting workspace where the agent is allowed to read invoices and bank statements and write reconciliation outputs. Run it against the intended clients directory, review generated JSON before relying on reminders or accounting decisions, and install poppler-utils from a trusted package source if pdftotext is missing.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Lp3

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding
Le skill prescrit explicitement l'exécution de `python3 scripts/main.py <racine_clients>` et la lecture/écriture de multiples fichiers (`followup.json`, `relances.json`, `anomalies.json`, rapports batch), ce qui implique des capacités shell, lecture et écriture non déclarées. L'absence de permissions explicites réduit la transparence et les garde-fous au moment de l'exécution, ce qui peut permettre à un agent d'utiliser des accès plus larges que prévu sur des données comptables sensibles.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.