ClawHub
Back to skill

Security audit

talkies

Security checks across malware telemetry and agentic risk

Overview

This skill does send speech and text to a configured Talkies server, but that behavior is clearly disclosed, purpose-aligned, and accompanied by practical safety guidance.

Install only if you trust the Talkies server you configure. Prefer localhost or a server you operate, enable TALKIES_AUTH_TOKEN for any shared or exposed deployment, use HTTPS or a trusted tunnel off-LAN, avoid sending secrets or regulated data, and clean up staged files and cached URL downloads when finished. Only use voice cloning with explicit speaker consent.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
Findings (3)

External Transmission

Medium
Category
Data Exfiltration
Content
# Discover TTS voices, then synthesize an MP3.
curl -s $TALKIES_URL/v1/audio/voices | jq
curl -s $TALKIES_URL/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{
        "model": "kokoro-82m",
Confidence
95% confidence
Finding
This TTS example sends arbitrary input text to an operator-configured remote service via $TALKIES_URL, which can exfiltrate sensitive user content if the endpoint is untrusted or misconfigured. The skill context makes this more significant because the documented purpose is to transmit audio/text to a speech server, and the server may also retain or further process that data.

External Transmission

Medium
Category
Data Exfiltration
Content
> ~/talkies-data/custom-voices/team-a/jane.txt

# Use it.
curl -s $TALKIES_URL/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{
        "model": "qwen3-tts-0.6b",
Confidence
97% confidence
Finding
This example transmits both synthesis text and a referenced cloned-voice sample to the remote Talkies server, creating a higher-risk data exposure path than ordinary TTS. In context, voice samples are especially sensitive biometric-like data, and the skill also notes shared server-side storage and optional auth, which increases misuse risk on shared or unauthenticated deployments.

External Transmission

Medium
Category
Data Exfiltration
Content
# Pick a voice from GET /v1/audio/voices, choose a format.
curl -s $TALKIES_URL/v1/audio/voices | jq -r '.voices[].voice'
curl -s $TALKIES_URL/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{
        "model": "kokoro-82m",
Confidence
95% confidence
Finding
This documented speech synthesis workflow transmits input text to an external endpoint controlled by TALKIES_URL, so confidential material could leave the host or trust boundary. The skill context partially mitigates this by openly warning about outbound transmission, but the capability remains real and security-relevant.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.