Back to skill

Security audit

Generate Healthcare Ad Creative Brief

Security checks for vulnerabilities and agentic risk

Overview

The skill appears to be a healthcare ad brief generator, but it asks for broad shell access that is not explained by its stated purpose.

Review this before installing because shell access is broader than needed for generating ad creative briefs. Prefer a version that removes Bash, and only grant Read if you expect the skill to use local campaign documents.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • 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
Findings (1)

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:4
Finding

Unnecessary Shell Capability Violates Least Privilege

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 4
Vulnerability Type: Excessive tool permissions
Risk Level: Medium

Complete Code Snippet:

yaml
---
name: Generate Healthcare Ad Creative Brief
description: "Plan campaign visuals and hooks for healthcare promotions. Use when working on paid campaign planning for healthcare operators, clinic teams, patient..."
allowed-tools: Bash, Read
metadata:

Technical Analysis

The Skill declares access to the arbitrary shell-execution tool Bash, although its documented purpose is limited to producing healthcare advertising creative briefs. The workflow contains no legitimate requirement to execute operating-system commands.

This violates the principle of least privilege. If attacker-controlled content, prompt injection, or future modifications influence the Skill, the unnecessary shell capability could become an execution channel. No malicious shell command or confirmed active exploitation is present in the audited file; the vulnerability is the excessive permission exposed by its configuration.

Attack Path

  1. The Skill is loaded and receives the tools declared under allowed-tools.
  2. The runtime grants it access to Bash.
  3. Attacker-controlled task content or injected instructions induce the agent to invoke shell commands.
  4. Those commands access host resources outside the legitimate healthcare brief-generation workflow.
  5. The resulting scope depends on the operating-system privileges, filesystem access, network access, and sandbox restrictions of the agent process.

Impact Assessment

Successful abuse could permit command execution with the privileges of the hosting agent process. Depending on runtime controls, this may expose readable local files, allow filesystem modification, invoke installed network utilities, or interact with other host resources. The issue does not independently demonstrate privilege escalation beyond ...[truncated 78 chars]

Remediation
View remediation

Remediation Suggestions

Remove Bash from allowed-tools because the documented workflow does not require shell execution:

yaml
allowed-tools: Read

Retain Read only if the Skill must consume local campaign materials. Otherwise, remove that permission as well. Explicitly allow only the chat and image-generation capabilities required by the stated workflow. If shell access later becomes necessary, expose narrowly scoped operations instead of unrestricted Bash and enforce sandboxing, filesystem boundaries, network restrictions, command allowlisting, and human approval for sensitive actions.

Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep