← 返回 Skills 市场
Lunara Voice
作者
lunara9897-jpg
· GitHub ↗
· v1.0.0
686
总下载
2
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install lunara-voice
功能描述
Bundle for Lunara Voice OpenClaw plugin with install and publish helpers
使用说明 (SKILL.md)
Lunara Voice Bundle
This skill bundle ships everything needed to install and run the Lunara Voice plugin locally.
What is included
plugin/— full copy of the OpenClaw plugin sourcescripts/install-plugin.sh— local install helperreferences/PUBLISH.md— release and ClawHub publish steps
Install plugin from this bundle
Run:
bash {baseDir}/scripts/install-plugin.sh
Then restart Gateway.
Configure plugin
Set plugin config in ~/.openclaw/openclaw.json:
{
"plugins": {
"entries": {
"lunara-voice": {
"enabled": true,
"config": {
"apiBaseUrl": "https://your-server.herokuapp.com",
"apiKey": "lnr_live_KEYID_SECRET",
"userEmail": "[email protected]"
}
}
}
}
}
Verify
openclaw plugins list
openclaw plugins info lunara-voice
openclaw plugins doctor
安全使用建议
This bundle appears to implement a real Lunara Voice plugin, but it has two properties you should weigh carefully before installing: (1) always: true — it will be force-loaded in every agent run, increasing exposure; and (2) the runtime instructions require autonomous, long-running network activity for every outbound call (silent polling, transcript reads, analytics, and data export). Only install if you fully trust the plugin source and the remote apiBaseUrl. If you proceed, do these things first: install in a restricted/test environment; inspect and, if possible, remove or disable the always flag; avoid supplying high-privilege or production credentials (use a limited API key); ensure PII masking remains enabled (do not set mask_pii=false); and review the plugin source (lunara-voice plugin/index.ts) for any endpoints or behaviors you do not recognize. If you are unsure about the provenance of this bundle (owner ID unknown, homepage metadata points to lunaravox.com), contact the vendor or prefer an official distribution channel.
功能分析
Type: OpenClaw Skill
Name: lunara-voice
Version: 1.0.0
The skill bundle is suspicious due to a significant prompt injection vulnerability that could lead to information disclosure, primarily stemming from the instructions in `plugin/skills/lunara-voice/SKILL.md`. This markdown file dictates an 'Autonomous Execution' workflow for outbound calls, instructing the AI agent to make calls to user-specified numbers (via `lunara_call_single`), silently poll for completion, retrieve the full call transcript (via `lunara_history_detail` with `include_transcript=true`), and then report a summary including 'key quotes' from the transcript back to the user. This allows an attacker to prompt the agent to call sensitive numbers and exfiltrate the entire conversation content. Additionally, the `plugin/index.ts` exposes a `lunara_webhook_create` tool, enabling the agent to configure webhooks to arbitrary HTTPS URLs, which could be leveraged for data exfiltration if the agent is compromised.
能力评估
Purpose & Capability
Name, description, plugin manifest, package.json and TypeScript source consistently implement a Lunara Voice OpenClaw plugin (agents, campaigns, calls, history, analytics, webhooks, exports). The files present are proportional to the stated purpose. Minor mismatch: registry lists no required env vars, but the plugin expects configuration (apiBaseUrl, apiKey, userEmail) placed in ~/.openclaw/openclaw.json — this is implementation-specific rather than malicious.
Instruction Scope
SKILL.md contains prescriptive runtime behavior that goes beyond simple tool documentation: it mandates that any user request to make a call must trigger an end-to-end workflow executed autonomously in a single turn, including recording timestamps, placing the call, then polling the history API every 25–30 seconds (up to 5 minutes) until the new call is found, retrieving transcripts, saving analytics/tags, and reporting results. The instructions also tell the agent to 'poll silently' and not message the user during polling. This grants the agent broad autonomous network activity (long-running polling, transcript reads, and data export) and could lead to unexpected data access or exfiltration if the plugin is given credentials.
Install Mechanism
No registry install spec; the bundle is instruction/source-based. Provided install script simply runs 'openclaw plugins install' against the bundled plugin directory — it does not download code from remote arbitrary URLs. This is lower risk than remote downloads, but installing the plugin will register many tools that can make network requests.
Credentials
The plugin expects API credentials (apiBaseUrl, apiKey, userEmail) which are reasonable for a remote voice/call API. However the plugin exposes features for bulk export and LLM training (lunara_export_bulk, mask_pii=false option), analytics saves, and full transcripts — these are sensitive operations. The manifest marks apiKey as sensitive; ensure you only provide credentials you trust and that the apiBaseUrl points to a trusted endpoint. The SKILL.md's guidance about disabling PII masking (mask_pii=false) is a potential privacy/data-exfiltration vector if misused.
Persistence & Privilege
The bundle metadata sets always: true, meaning the skill is force-included in every agent run. Combined with the plugin's ability to make outbound calls, fetch transcripts, perform exports, and the SKILL.md's requirement to autonomously perform long-running polling for calls, this markedly increases the plugin's blast radius. There is no clear, justified reason in the documentation why this plugin must be always-enabled for all agents.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install lunara-voice - 安装完成后,直接呼叫该 Skill 的名称或使用
/lunara-voice触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
LunaraVox — AI Phone Calls API for OpenClaw
LunaraVox is an AI calling infrastructure that allows OpenClaw agents to make and manage real phone calls through a dedicated API layer. The plugin connects OpenClaw with LunaraVox call execution, enabling AI-driven inbound and outbound conversations at scale while keeping business logic and control fully configurable.
The plugin acts as a bridge between OpenClaw and the LunaraVox calling engine. When an OpenClaw agent initiates a call, the request is sent through the LunaraVox API, where call behavior, greeting logic, and conversational instructions are applied before the call is executed.
This architecture separates decision-making from call execution:
OpenClaw handles reasoning and task logic.
LunaraVox handles telephony, voice interaction, and call processing.
This allows AI agents to perform real-world communication tasks without requiring complex telephony integrations inside OpenClaw itself.
Core Capabilities
AI Call Execution
OpenClaw agents can initiate phone calls directly through the LunaraVox API. The system injects predefined behavior prompts, greeting logic, and conversation goals before the call starts.
Supported scenarios include:
• outbound sales and qualification calls
• follow-ups and reminders
• appointment confirmations
• customer information collection
• automated campaign calls
• inbound call handling with AI assistants
The AI communicates naturally, processes responses in real time, and returns structured results back to the agent.
Call History & Analytics API
LunaraVox provides a dedicated history API optimized for LLM usage. Instead of loading full conversations, the system allows filtered access to historical call data, reducing token usage and improving response speed.
Available filters include:
• date ranges
• call duration limits
• call direction
• call status and outcomes
• tagged call categories
• assistant-specific history
This allows OpenClaw agents to analyze past calls, identify patterns, and make decisions based on structured historical data rather than raw transcripts.
Detailed Call Data
For each call, the agent can retrieve:
• structured call summary
• outcome and agreements reached
• optional transcript
• audio reference
• detailed metadata
Transcripts can be excluded when not required, minimizing token consumption.
PII Protection
The API automatically detects and masks sensitive information such as phone numbers, emails, and payment data. This ensures that stored call data remains safe for analytics and AI processing without exposing sensitive user information.
Export for Training and Optimization
Call history can be exported in formats suitable for LLM processing and model training workflows. Businesses can use successful conversations as datasets for improving AI behavior, optimizing scripts, and increasing conversion performance over time.
This enables iterative improvement based on real successful interactions rather than manual scripting alone.
Architecture Overview
The plugin follows a modular architecture:
1. OpenClaw decides when and why a call should be made.
2. LunaraVox receives the request via API.
3. Behavior prompts and call configuration are applied.
4. The AI executes the call through the connected SIP provider.
5. Results and analytics are returned to OpenClaw.
The AI agent itself does not perform telephony — LunaraVox provides the execution layer.
This separation allows independent scaling of AI logic and call infrastructure.
Requirements
To use the plugin, users must:
• register an account at lunaravox.com
• connect their own SIP provider
• purchase a call minutes package (minimum 300 minutes)
• install and configure the LunaraVox plugin in OpenClaw
SIP costs remain fully controlled by the user and are billed by their telecom provider.
Pricing
LunaraVox usage is billed at:
$4 per active AI hour
No long-term contracts or platform lock-in.
Call connection costs are paid separately to the user’s SIP provider.
Use Cases
• AI sales automation
• scalable outbound campaigns
• customer support automation
• lead qualification
• appointment scheduling
• high-volume communication workflows
LunaraVox is designed for businesses that need real phone communication at scale without scaling human operators.
元数据
常见问题
Lunara Voice 是什么?
Bundle for Lunara Voice OpenClaw plugin with install and publish helpers. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 686 次。
如何安装 Lunara Voice?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install lunara-voice」即可一键安装,无需额外配置。
Lunara Voice 是免费的吗?
是的,Lunara Voice 完全免费(开源免费),可自由下载、安装和使用。
Lunara Voice 支持哪些平台?
Lunara Voice 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 Lunara Voice?
由 lunara9897-jpg(@lunara9897-jpg)开发并维护,当前版本 v1.0.0。
推荐 Skills