Skills tts-skill

📦

tts-skill

Name: tts-skill
Author: tabortao

Content revision r1 Critical ⚙️ External commands🌐 Network access🔑 Env variables📁 Filesystem access

Generate Speech With Multiple TTS Engines

Creators need speech output across local and online voice engines. This skill gives Claude, Codex, and Claude Code one workflow for Qwen3, Edge-style, and OpenAI TTS.

Supports: Claude Codex Code(CC)

⚠️ 38 Poor

Do not auto-install this skill.

The canonical policy requires operator review before any installation action.

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "tts-skill" from https://skillstore.io/skills/tabortao-tts-skill.md and its manifest at https://skillstore.io/api/skills/tabortao-tts-skill/manifest. Verify the artifact. Do not auto-install. Inspect the skill and report your findings, then wait for an operator or manual installation decision.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/tabortao-tts-skill.md Signed manifest GET /api/skills/tabortao-tts-skill/manifest Signed lockfile GET /api/skills/tabortao-tts-skill/lockfile

Test it

Using "tts-skill". Generate a welcome line with an online Chinese voice.

Expected outcome:

The skill selects the Edge-style online engine, sends the text to the configured endpoint, and saves an MP3 file.

Using "tts-skill". Create a cloned voice sample from a local reference voice.

Expected outcome:

The skill locates matching reference audio and transcript files, runs the Qwen3 engine, and writes a WAV output with timing details.

Using "tts-skill". List the voices I can use before generating audio.

Expected outcome:

The skill prints local voice names from assets and built-in online voice aliases for Edge-style and OpenAI engines.

Security Audit

Critical

v4 • 7/7/2026 Open versioned report

The audit confirms real external command execution, outbound TTS API calls, filesystem writes, and API-key handling. The most serious issue is a critical Qwen3 generated-code injection path where user text is inserted into Python source and executed. No evidence found of prompt-injection instructions in the reviewed skill files.

Files scanned

2,706

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (2)

Critical

docs/INSTALL.zh-CN.md:96

Pipe to shell pattern

curl -Ls https://micro.mamba.pm/install.sh | sh

The command pipes a remote install script directly to sh. This is a classic unsafe installer pattern because network content executes without local review.

Critical

engines/qwen3-tts-cli.py:175-211 engines/qwen3-tts-cli.py:346-356

Generated Python Source Injection

Qwen3-TTS builds a temporary Python source file that inserts raw user text into executable code, then writes and runs that file. Malicious input text can execute arbitrary Python commands in the micromamba environment.

The source template concatenates raw text into a Python print statement, then writes temp_qwen3_generate.py and executes it. This is direct code generation from user-controlled input.

Capability review items (31)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Critical

docs/INSTALL.zh-CN.md:96

Hardcoded URL

curl -Ls https://micro.mamba.pm/install.sh | sh

The URL is used in a curl pipe-to-shell install command. Remote script contents would execute immediately in the user environment.

Critical

engines/qwen3-tts-cli.py:144

Python subprocess.run

subprocess.run(['powershell', '-Command', ps_command], check=True)

The auto-installer executes a remote Micromamba PowerShell install command through powershell -Command. Remote installer code runs without verification.

Critical

engines/qwen3-tts-cli.py:356

Python subprocess.run

result = subprocess.run(cmd, env=env, cwd=str(engines_dir))

This executes a generated temporary Python script whose source includes user text. That creates an arbitrary Python code execution path.

Critical

engines/qwen3-tts-cli.py:144

PowerShell invocation

subprocess.run(['powershell', '-Command', ps_command], check=True)

The command runs PowerShell with Invoke-Expression on downloaded installer content. This is a high-risk remote code execution pattern.

Critical

engines/qwen3-tts-cli.py:143

Hardcoded URL

ps_command = "Invoke-Expression ((Invoke-WebRequest -Uri https://micro.mamba.pm/install.ps1 -UseBasi

The hardcoded URL is embedded in a PowerShell Invoke-Expression installer command. The installer content can execute in the user environment.

Critical

engines/qwen3-tts-cli.py:346

Python file write/append

with open(temp_script, 'w', encoding='utf-8') as f:

The code writes generated Python source to a temporary file before executing it. That source includes unsafely concatenated user text.

High

docs/INSTALL.zh-CN.md:68

PowerShell invocation

```powershell

The surrounding install block instructs Windows users to run a remote Micromamba PowerShell installer with Invoke-Expression. Downloaded installer code runs without checksum or signature verification.

High

docs/INSTALL.zh-CN.md:194

PowerShell invocation

- Windows: 使用PowerShell运行安装脚本

The troubleshooting text directs users to run a PowerShell install script. This is manual documentation, but it still normalizes remote installer execution.

High

docs/INSTALL.zh-CN.md:73

Hardcoded URL

Invoke-Expression ((Invoke-WebRequest -Uri https://micro.mamba.pm/install.ps1 -UseBasicParsing).Cont

The URL is used in a PowerShell Invoke-Expression installer command in the same block. Users are instructed to fetch and execute internet code without verification.

High

engines/openai-tts-cli.py:145

Python HTTP libraries

response = requests.post(

The code posts user text to the configured OpenAI-compatible speech endpoint. It also attaches an Authorization bearer token in the same request.

High

engines/openai-tts-cli.py:71

Configuration library

self.api_key = openai_config.get('api_key', '')

This line reads api_key from the OpenAI configuration. It is a real credential-handling path, even though it is expected for this engine.

High

engines/openai-tts-cli.py:71

Generic API/secret keys

self.api_key = openai_config.get('api_key', '')

The code reads api_key from the configuration into memory. This is a real credential-handling path for the OpenAI engine.

High

engines/openai-tts-cli.py:148

Generic API/secret keys

'Authorization': f'Bearer {self.api_key}',

The API key is placed in an Authorization bearer header. If base_url is changed, the credential can be sent to a non-OpenAI endpoint.

High

engines/qwen3-tts-cli.py:146

Python subprocess.run

subprocess.run(['curl', '-Ls', 'https://micro.mamba.pm/install.sh'],

The auto-install path invokes curl against a remote installer URL. The arguments are fixed, but automatic remote installer retrieval still creates supply-chain risk.

High

engines/qwen3-tts-cli.py:151

Python subprocess.run

subprocess.run(['micromamba', 'create', '-n', 'qwen3-tts', 'python=3.12', '-y'], check=True)

The script creates a micromamba environment from inside the skill. This changes local environment state and executes external tooling.

High

engines/qwen3-tts-cli.py:155

Python subprocess.run

subprocess.run(['micromamba', 'run', '-n', 'qwen3-tts', 'pip', 'install', '-U', 'qwen-tts'], check=T

The script installs qwen-tts at runtime through pip. Runtime package installation executes third-party package code and adds supply-chain risk.

High

engines/qwen3-tts-cli.py:159

Python subprocess.run

subprocess.run(['micromamba', 'run', '-n', 'qwen3-tts', 'pip', 'install', '-U', 'modelscope'], check

The script installs modelscope at runtime through pip. Runtime package installation executes third-party package code and adds supply-chain risk.

High

engines/qwen3-tts-cli.py:146

Hardcoded URL

subprocess.run(['curl', '-Ls', 'https://micro.mamba.pm/install.sh'],

The script fetches a remote Micromamba installer URL during automatic setup. The lack of checksum or signature verification is a real supply-chain risk.

High

tts-skill.py:167

Python subprocess.run

result = subprocess.run(cmd, cwd=str(self.engines_dir),

The main entrypoint runs selected engine scripts as subprocesses and forwards user arguments. Engine selection is allowlisted, but the child engines perform network and file operations.

High

tts-skill.py:189

Python subprocess.run

result = subprocess.run(cmd, cwd=str(self.engines_dir))

The install path invokes the Qwen engine installer as a subprocess. That installer can download tools and install packages into the local environment.

Medium

docs/INSTALL.zh-CN.md:274

Hardcoded URL

api_url = https://backup-tts-api.com/v1/audio/speech

The documentation suggests a backup TTS API endpoint that could receive user text. The endpoint is not an official provider in the reviewed context.

Medium

engines/edge-tts-cli.py:147

Python HTTP libraries

response = requests.post(

The code posts user-provided text and voice parameters to the configured Edge-style TTS endpoint. That is a real outbound data transfer.

Medium

engines/edge-tts-cli.py:37

Hardcoded URL

'api_url': 'https://tts.wangwangit.com/v1/audio/speech',

The default endpoint points to an external TTS service. Text submitted to this engine leaves the local environment.

Medium

engines/edge-tts-cli.py:174

Python file write/append

with open(output_path, 'wb') as f:

The engine writes the downloaded audio response to the requested output path. A user-supplied path can overwrite files allowed by process permissions.

Medium

engines/edge-tts.config:3

Hardcoded URL

api_url = https://tts.wangwangit.com/v1/audio/speech

The config sets an external TTS API endpoint as the default. User text can be sent to that service when the engine runs.

Medium

engines/openai-tts-cli.py:72

Hardcoded URL

base_url = openai_config.get('base_url', 'https://api.openai.com/v1')

The client builds its speech endpoint from a configurable base URL, defaulting to OpenAI. Generated text will be sent to that external service.

Medium

engines/openai-tts-cli.py:165

Python file write/append

with open(output_path, 'wb') as f:

The engine writes the API response to the requested output path. A user-supplied path can overwrite files allowed by process permissions.

Medium

engines/qwen3-tts-cli.py:350

Python environment access

env = os.environ.copy()

The generated child process receives a copy of the full parent environment. Any inherited secrets are exposed to imported packages and injected code.

Medium

tts-skill.py:169

Python environment access

env={**os.environ, 'PYTHONIOENCODING': 'utf-8', 'PYTHONUTF8': '1'})

The main wrapper passes the full inherited environment to engine subprocesses. Secrets present in the environment become available to child code.

Low

engines/openai-tts-cli.py:58

Hardcoded URL

'base_url': 'https://api.openai.com/v1',

The default configuration uses the OpenAI API base URL. This is expected for the engine, but it is a real external network dependency.

Low

engines/openai-tts.config:11

Hardcoded URL

base_url = https://api.openai.com/v1

The default config points to the OpenAI API service. This is expected, but it confirms an external network dependency for the OpenAI engine.

Risk Factors

⚙️ External commands (38)

🌐 Network access (20)

🔑 Env variables (27)

📁 Filesystem access (9)

engines/edge-tts-cli.py:174 engines/openai-tts-cli.py:165 engines/qwen3-tts-cli.py:63 engines/qwen3-tts-cli.py:346 engines/qwen3-tts-cli.py:361 engines/qwen3-tts.config:16 engines/qwen3-tts.config:34 engines/qwen3-tts.config:49 README.md:48

Detected Patterns

Pipe to shell pattern

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/tabortao-tts-skill/audits/4?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/tabortao-tts-skill/security.svg)](https://skillstore.io/skills/tabortao-tts-skill?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/tabortao-tts-skill?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/tabortao-tts-skill/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/tabortao-tts-skill.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

tabortao. (2026). tts-skill security audit report (audit version 4) [Author version unspecified]. Skillstore. https://skillstore.io/skills/tabortao-tts-skill/audits/4

BibTeX citation

@techreport{tabortao-tabortao-tts-skill-2026,
  author = {tabortao},
  title = {tts-skill security audit report (audit version 4)},
  institution = {Skillstore},
  year = {2026},
  number = {4},
  url = {https://skillstore.io/skills/tabortao-tts-skill/audits/4},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "tts-skill security audit report (audit version 4)"
version: "unspecified"
type: report
authors:
  - name: "tabortao"
date-released: "2026-07-07"
url: "https://skillstore.io/skills/tabortao-tts-skill/audits/4"
identifiers:
  - type: other
    value: "skillstore:tabortao-tts-skill:audit:4"
    description: "Skillstore immutable audit report identifier"

Skillstore Score

Why this score Evidence Confidence: Medium

Architecture

Maintainability

Content

Community

Spec Compliance

What You Can Build

Prototype Narration Audio

Create quick speech samples from draft copy using online voices before recording final audio.

Generate Local Character Voices

Use Qwen3 reference assets to test cloned voices for game dialogue or product demos.

Convert Script Files

Read long script text from files and save timestamped speech outputs for review workflows.

Try These Prompts

Create a Simple Voice Clip

Use tts-skill to generate a short speech clip from this sentence with the xiaoxiao voice: Welcome to the product tour.

Use a Text File

Use tts-skill with the Qwen3-TTS engine. Read the approved input text file, use the Lei voice, and save the audio in the output folder.

Compare Voice Engines

Generate the same short paragraph with Edge-style TTS and OpenAI TTS. Use clear voice settings and report the output files created.

Prepare a Controlled TTS Run

Review the configured TTS engine, confirm network and credential requirements, then generate speech from my approved script with an explicit output path.

Best Practices

Use explicit output paths inside a dedicated output directory.
Review text before sending it to any online TTS provider.
Store real API keys outside shared repositories and rotate them regularly.

Avoid

Do not run the Qwen3 engine on untrusted text until generated-code execution is fixed.
Do not pipe remote installer scripts to a shell without independent verification.
Do not place private scripts, secrets, or regulated data in online TTS requests.