ClawHub
Back to skill

Security audit

Call Report Writer Agent

Security checks for vulnerabilities and agentic risk

Overview

The skill is not clearly malicious, but its name and description understate a much broader automated research, code-execution, scraping, visual-generation, and submission workflow.

Review this skill carefully before installing. It should be treated as a broad multimedia research/report-generation agent, not a narrow call report writer; only use it where web access, screenshot capture, Python execution, document creation, and automatic result submission are intended and appropriately permissioned.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (3)

Description-Behavior Mismatch

Medium
Confidence
97% confidence
Finding
The skill metadata says it is a call report writer, but the embedded instructions redefine it into a broad autonomous research-and-publication agent with external URL investigation, screenshot capture, code-driven chart generation, and document submission. This capability mismatch is dangerous because users or orchestration layers may grant it permissions appropriate for a narrow reporting task while the skill actually attempts much wider data access and output actions.

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The instructions create an intent-level contradiction by labeling the skill as a call report writer while directing it to behave as a 'Multimedia Research Producer' producing media-rich investigative reports. This inconsistency increases the chance of deceptive tool enablement, policy bypass by misclassification, and operator misunderstanding about what the skill will actually do.

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
The skill instructs the agent to perform networked research, scrape URLs, capture screenshots, and execute Python to generate assets, even though those capabilities are not clearly necessary for a call report writer. In practice, this broadens the attack surface for prompt injection, data exfiltration through external requests, and unsafe use of privileged tools under an innocuous-looking skill name.