T01 · Skill Instruction Hijacking
- Location
SKILL.md:92- Finding
Persistent Agent Behavior Hijacking Through SOUL.md Modification
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This TTS skill is not clearly malicious, but it asks to persistently force all future messages through background speech generation and package installation with weak user control.
Install only if you intentionally want persistent automatic speech for all agent messages. Review and remove the SOUL.md instruction, avoid global or automatic npm installs, pin/review dependencies, and confirm that the missing auto-speak executable is supplied from a trusted source before use.
SKILL.md:92Persistent Agent Behavior Hijacking Through SOUL.md Modification
package.json:17Unsafe Installation of a Mutable Third-Party npm Dependency
package.json:4Package Declares a Missing Executable and May Invoke an Unrelated PATH Command
The skill claims every message is automatically spoken using Microsoft Edge's neural TTS service, which implies conversation content is transmitted to an external provider, but it does not clearly warn users about that data flow or its privacy implications. Because this applies to every message by default and runs automatically in the background, users may unknowingly expose sensitive prompts, secrets, or personal data to a third party.
The configuration hard-codes a specific locale voice (en-US-MichelleNeural) as the default without explicit user opt-in, which can impose language, accent, or regional assumptions on all spoken output. In this skill's context, the issue is less about code execution and more about unauthorized behavior and accessibility/privacy expectations, since automatic TTS of every message amplifies the effect of any unwanted default.
The config hard-codes the voice to "en-US-MichelleNeural", which imposes a specific language and locale. Under the policy, locale constraints should not be forced unless the skill offers user choice or clearly documents a justified region-specific requirement.
The dependency uses a caret range (^1.0.0), which permits automatic installation of newer minor and patch releases that have not been explicitly reviewed. In a skill that autonomously installs and executes a TTS-related package, this increases supply-chain risk because a compromised upstream release could be pulled into the environment without deliberate pinning or integrity review.
"author": "Stefano",
"license": "MIT",
"dependencies": {
"node-edge-tts": "^1.0.0"
},
"bin": {
"auto-speak": "./auto-speak"