Security audit

Stego Text — Steganographic Message Encoding for AI Agents

Security checks for vulnerabilities and agentic risk

Overview

The skill is text-only, but it teaches hard-to-detect hidden messaging without clear safety limits.

Review before installing. This skill has no code-level access risk, but it is designed to create and analyze concealed messages. Use it only for clearly benign contexts such as puzzles, watermarking research, or defensive analysis, and avoid using it to bypass moderation, hide instructions, or conceal sensitive data movement.

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

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The skill description contains broad triggers such as 'hide a message', 'encode a secret message', and 'analyze text for hidden encodings', which can activate on general requests and route users into covert-communication functionality without tight scoping. In the context of a steganography skill, this increases misuse risk because the core purpose is concealment, making accidental or inappropriate invocation more dangerous than in a benign formatting or educational skill.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The documentation teaches multiple covert text-hiding methods and emphasizes properties like 'very hard' detection difficulty, but provides no warning, restriction, or safety boundary regarding misuse. That omission materially increases risk because users are given operational guidance for concealed communication without any notice about compliance, abuse, or the need to refuse harmful evasion-oriented requests.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: This file explicitly teaches multiple covert communication and steganographic techniques, including methods framed as difficult for humans to notice and detectable mainly through statistical analysis. In an agent skill, that materially enables misuse for hidden instruction passing, policy evasion, or exfiltration, and the absence of warnings, abuse boundaries, or defensive framing increases the risk.

Static analysis

No suspicious patterns detected.