Back to skill

Security audit

Generate Customer Support Ad Creative Brief

Security checks for vulnerabilities and agentic risk

Overview

The skill is a simple ad-brief generator, but it requests shell command access that is not explained or needed for that purpose.

Review this skill before installing. Its content-generation behavior is straightforward, but it should not need shell command access; prefer a version that removes Bash from allowed tools unless you have a specific, trusted reason to let it run local commands.

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 Bash Permission Violates Least Privilege

Content
View full analysis

Vulnerability Details

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

Vulnerable Code Snippet:

yaml
allowed-tools: Bash, Read

Technical Analysis

The skill grants access to Bash, although its documented purpose is limited to preparing advertising creative briefs. The documented workflow consists of clarifying campaign requirements, generating content, enriching assets, and refining the output; it does not identify any legitimate need to execute operating-system commands.

Bash access allows commands to run with the permissions of the account hosting the agent. This unnecessarily expands the skill's authority and violates the principle of least privilege. The declaration does not itself execute commands and no malicious command is embedded in the reviewed file. Nevertheless, the excessive capability could turn malicious or prompt-injected instructions into local command execution.

Attack Path

  1. An attacker introduces crafted instructions through campaign requirements, referenced content, or another untrusted input processed while this skill is active.
  2. The crafted content instructs the agent to invoke Bash under the guise of gathering supporting data or enriching campaign assets.
  3. Because Bash is explicitly permitted, the agent may execute the injected command.
  4. The command could access files, alter data, invoke installed programs, or communicate with external systems, subject to the hosting process's operating-system privileges and any external sandbox restrictions.

This path requires the agent to follow attacker-controlled instructions; the audited file contains no direct command invocation or confirmed exploitation logic.

Impact Assessment

Successful exploitation could provide command execution within the security context of the agent process. Potential scope includes reading files available to that account, modifying ...[truncated 265 chars]

Remediation
View remediation

Remediation Suggestions

Remove Bash from allowed-tools because no shell operation is required by the documented workflow:

yaml
allowed-tools: Read

Also remove Read if the skill does not need to process local files. If future functionality genuinely requires command execution, replace unrestricted Bash access with a narrowly scoped tool that exposes only the required operation. Validate untrusted campaign content before using it in tool calls, require explicit user approval for consequential actions, and enforce runtime sandboxing with minimal filesystem and network permissions.

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