Back to skill

Security audit

Generate Real Estate Ad Creative Brief

Security checks for vulnerabilities and agentic risk

Overview

This real-estate ad-brief skill is mostly straightforward, but it requests broad Bash command access that does not fit its writing-focused purpose.

Install only if you are comfortable giving this skill shell-command capability in your agent environment. For normal real-estate ad-brief drafting, Bash should be removed or tightly scoped; review generated briefs before publishing or automating any campaign work.

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:8
Finding

Unnecessary Bash Permission Violates Least Privilege

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:8
Vulnerability Type: Excessive tool permission
Risk Level: Medium

yaml
allowed-tools: Bash, Read

Technical Analysis

The skill grants unrestricted Bash access even though its declared workflow only involves clarifying advertising requirements, drafting a real-estate creative brief, enriching supporting materials, and refining the result. No documented workflow step requires shell command execution.

This permission violates the principle of least privilege. While the reviewed skill contains no malicious shell command, granting Bash makes a command-execution capability available during skill operation. Malicious or untrusted task content could attempt to induce the agent to use that capability outside the legitimate creative-brief workflow.

Attack Path

  1. A user or system loads the skill, activating its declared Bash permission.
  2. The agent processes attacker-controlled or otherwise untrusted campaign content.
  3. That content attempts to persuade the agent to invoke Bash under the pretext of collecting assets, enriching supporting data, or completing the brief.
  4. If the induced tool call is accepted, shell commands execute with the operating-system privileges and sandbox access assigned to the agent process.
  5. The attacker may consequently access files, inspect the execution environment, or perform other commands permitted by that runtime.

This is a capability-enablement path rather than evidence of active exploitation: no shell invocation or malicious payload was found in the reviewed files.

Impact Assessment

Successful exploitation could provide command execution within the agent's runtime security boundary. The practical scope depends on external sandboxing and process privileges, potentially including readable or writable local files, environment information, network resources, and executable programs available to the agent. The rev ...[truncated 115 chars]

Remediation
View remediation

Remediation Suggestions

Remove Bash from allowed-tools, because the documented workflow does not require command execution. Remove Read as well unless processing local source documents is an explicit and necessary use case.

Define the smallest possible tool allowlist and document why each retained capability is needed. If future functionality genuinely requires shell operations, replace unrestricted Bash access with a narrowly scoped tool that validates commands, arguments, file paths, and network destinations. Run such operations in a restricted sandbox with minimal filesystem permissions, no unnecessary credentials, controlled network access, execution timeouts, and audit logging.

Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The markdown includes a suggested prompt, "[Generate Real Estate Ad Creative Brief] for my team," but does not provide any negative examples or clear boundaries for when the skill should or should not be invoked. While the skill name is somewhat specific, the invocation guidance remains underspecified and could cause unintended activation in general team-planning contexts.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.