Back to skill

Security audit

real-estate-listing-writer

Security checks for vulnerabilities and agentic risk

Overview

The skill uses a limited local drafting workflow, though its helper produces an outline rather than finished, checked listing copy.

Treat the script output as a preliminary outline, and independently check facts and wording before publication. Use a dedicated output folder because outline.md is overwritten. The optional Windows app linked through the [publisher repository](https://github.com/rapidworker1/Multiple-AI-models-Toolkit) is unnecessary for this workflow and requires a separate security review.

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
Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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
Findings (4)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The primary output materially differs from the declared purpose: the script produces a generic keyword outline, not property listing copy based on a seller fact sheet. Although its static guardrails align with the declared subject matter, they do not implement a compliance or accuracy review. Local input reading and output writing are ordinary supporting operations; no unrelated resource access or hidden capability is evident.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_write.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The script generates a keyword-based content outline rather than MLS-ready listing copy grounded in a seller's fact sheet.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The manifest describes writing property descriptions from seller-provided facts, but this implementation consumes only the keywords field and emits generic section prompts such as 'answer the query' and 'add one example and one US-specific detail' (L073–L084). Fair-housing and accuracy limits are appended as reminders (L085–L087), rather than applied to generated listing copy; the resulting artifact is an outline, not the described MLS-ready description.

Content

No source excerpt is available for this finding.

The generated outline requires a US-specific detail without user opt-in or a documented regional scope.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

Every keyword section is instructed to include a US-specific detail, regardless of the property's location or the user's locale. The file describes a general real estate listing writer and provides neither a user locale choice nor justification for restricting content to the US.

Content

No source excerpt is available for this finding.