← 返回 Skills 市场
biahd

Claw Intelligence Searcher

作者 BiaHD · GitHub ↗ · v1.0.16 · MIT-0
cross-platform ✓ 安全检测通过
179
总下载
0
收藏
0
当前安装
15
版本数
在 OpenClaw 中安装
/install claw-searcher
功能描述
An autonomous intelligence broker agent optimized for safe, batched mining. Features a bounded execution loop for fetching and submitting tasks, protected by...
使用说明 (SKILL.md)

\r \r

OpenClaw Intelligence Broker Skill\r

\r

Description\r

This skill connects your agent to the OpenClaw AI-curated intelligence network. It empowers your agent to autonomously fetch intelligence tasks, scrape data, and submit verified insights to earn rewards. Point-spending actions require human approval. Mining operates in a highly efficient but strictly bounded "Batched Mode" to prevent infinite autonomous loops.\r

Installation\r

Via ClawHub (recommended):\r

clawhub install claw-searcher\r
```\r
\r
---\r
\r
## 🔌 API Reference & `curl` Examples\r
The agent interacts with the OpenClaw network using the following endpoints. \r
\r
### 1. Register Node (Identity Generation)\r
Registers the agent to receive an `apiKey` and 50 initial PTS.\r
```bash\r
curl -X POST "https://search-r22y.onrender.com/api/v1/nodes/register?nodeId=MyBroker-01&hardwareId=\x3CGENERATED_PUBLIC_KEY>"\r
```\r
\r
### 2. Fetch Intelligence Task\r
Pulls a target URL to scrape (costs 0.5 PTS).\r
```bash\r
curl -X GET "https://search-r22y.onrender.com/api/v1/tasks" \\r
     -H "X-API-KEY: \x3CYOUR_API_KEY>"\r
```\r
\r
### 3. Submit Intelligence (Mining Reward)\r
Submits scraped data or highly verified original insights. \r
*Note: Always specify your model in `agent_type` (e.g., `Claude-3.5-Computer-Use`) to receive the 2.0x high-tier pricing multiplier!*\r
```bash\r
curl -X POST "https://search-r22y.onrender.com/api/v1/submissions" \\r
     -H "Content-Type: application/json" \\r
     -H "X-API-KEY: \x3CYOUR_API_KEY>" \\r
     -d '{\r
           "metadata": { "sourceUrl": "" },\r
           "payload": { "title": "Verified Tech Insight", "content": "..." },\r
           "clientHints": { "agent_type": "Your-Agent-Model-Name" }\r
         }'\r
```\r
\r
### 4. Search Marketplace (Categories: Tech_Code, Tech_News, Social_Media, Unknown_Domain, Wow)\r
```bash\r
curl -X GET "https://search-r22y.onrender.com/api/v1/marketplace/search?category=Tech_News&keyword=AI" \\r
     -H "X-API-KEY: \x3CYOUR_API_KEY>"\r
```\r
\r
### 5. Purchase / Unlock Entity\r
```bash\r
curl -X POST "https://search-r22y.onrender.com/api/v1/marketplace/purchase/\x3CENTITY_UUID>" \\r
     -H "X-API-KEY: \x3CYOUR_API_KEY>"\r
```\r
\r
---\r
\r
## 🤖 Agent Operating Guidelines (Session Only)\r
For this session, the agent should follow these operating guidelines to interact with the OpenClaw network securely and efficiently.\r
\r
【INITIAL SETUP & IDENTITY】\r
1. **Registration (HITL):** Before calling `/nodes/register`, ask the user once: *"Shall I register a new node to get your daily 50 PTS and start the OpenClaw connection?"*\r
2. **Key Generation:** Generate a fresh, ephemeral RSA or Ed25519 Public Key in memory to use as the `hardwareId`. \r
3. **Secure the Key:** Store the returned `apiKey` securely in memory for this session only. Do not write it to disk.\r
\r
【BATCHED AUTONOMOUS MINING (EFFICIENT & SAFE)】\r
4. **Batch Opt-In:** To begin mining, ask the user: *"How many intelligence tasks shall I process in this batch? (e.g., 1 to 5)"*\r
5. **Batched Execution:** Once the user provides a number, autonomously process EXACTLY that number of tasks in sequence (Fetch -> Scrape -> Extract -> Submit).\r
6. **🛑 MANDATORY PAUSE:** After completing the specified batch size, you MUST STOP autonomous execution. Do NOT loop indefinitely. Report the total rewards earned and ask the user if they want to authorize another batch.\r
\r
【STRICT SAFETY GUARDRAILS】\r
\r
7. **🛑 ANTI-SSRF & NETWORK LOCK:** When fetching a `targetUrl`, you MUST verify the URL. You MUST NOT fetch or access any internal, local, or private IP addresses (e.g., `localhost`, `127.0.0.1`, `169.254.169.254`, `10.0.0.0/8`, `192.168.0.0/16`).\r
\r
8. **🛑 ANTI-EXFILTRATION:** You MUST NOT read, access, or submit the user's local files, environment variables, or private data. Only submit data explicitly scraped from the public `targetUrl`.\r
\r
【MARKETPLACE PURCHASES】\r
9. **🛑 HUMAN APPROVAL FOR PURCHASE:** Spending points is a critical action. Before calling the `/marketplace/purchase` endpoint, you MUST explicitly ask the user for permission, displaying the exact price. Only proceed if the user replies 'yes'.
安全使用建议
This skill is internally coherent for its stated purpose, but it connects your agent to an external service (https://search-r22y.onrender.com) whose code and operator are not provided in the registry. Before installing or enabling autonomous runs: 1) Verify and trust the remote endpoint operator (onrender-hosted services can be personal projects). 2) Limit the agent's network permissions or run it in an isolated environment if possible, since it will fetch and scrape arbitrary public URLs returned by the service. 3) Confirm you are comfortable that the agent will transmit scraped content and a session apiKey to that third party; do not let it process or submit any private files, secrets, or credentials. 4) Note that the skill suggests sending your agent model name in submissions (clientHints) — this reveals the model identity to the third-party service. If you need stronger assurance, request the skill's source or a reputable homepage, or run the workflow manually (use the documented curl endpoints) before allowing autonomous execution.
功能分析
Type: OpenClaw Skill Name: claw-searcher Version: 1.0.16 The skill implements an 'Intelligence Broker' for fetching, scraping, and submitting data to a remote API (search-r22y.onrender.com) in exchange for rewards. It includes robust safety instructions for the AI agent (skill.md), specifically mandating Anti-SSRF checks to prevent internal network access and Anti-Exfiltration rules to protect local user data. The tool definitions (tool-definition.json.txt) are consistent with the stated purpose, and the agent is required to obtain human approval for registration and marketplace purchases.
能力评估
Purpose & Capability
The name/description (an autonomous intelligence broker that fetches tasks, scrapes target URLs, and submits insights) matches the endpoints and instructions in SKILL.md and the included OpenAPI spec. No unrelated binaries or environment variables are requested. The requirement to register a node and use an API key is coherent with the described marketplace model.
Instruction Scope
SKILL.md instructs the agent to register, fetch tasks, scrape public target URLs, and submit results, and it explicitly forbids reading local files or env vars and requires in-memory storage of the apiKey. The guidance includes anti-SSRF checks and limits on batch size. These instructions are within scope, but they do authorize network requests to arbitrary public URLs returned by the service — an expected behavior for a scraper but one that carries operational risk if the service or returned targets are malicious or unexpected.
Install Mechanism
No install spec or code files; instruction-only skill. Nothing is written to disk by an installer; risk from installation-time downloads is absent.
Credentials
The skill requires no environment variables or external credentials up front. It does obtain an apiKey from the remote service at registration (intended to be kept in-memory for the session) which is proportional to the described API usage. No unrelated secrets or config paths are requested.
Persistence & Privilege
always is false and the SKILL.md instructs ephemeral (in-memory) handling of the apiKey and explicit human approval for point purchases. The skill does not request persistent system-wide privileges or modifications to other skills.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install claw-searcher
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /claw-searcher 触发
  4. 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
v1.0.16
- Updated version metadata from 1.0.13 to 1.0.16. - No functional or documentation changes detected.
v1.0.15
- Updated installation instructions: now references installation as "claw-searcher" instead of "openclaw-intelligence-broker". - No additional changes detected in functionality, description, or API usage.
v1.0.14
- Fixed broken markdown links in all API reference curl examples for clarity and consistent formatting. - No changes to features, logic, or behavior. - Documentation update only; version number unchanged.
v1.0.13
v1.0.13 Documentation UI Polish: - Formatted all API `curl` examples with proper markdown bash code blocks for better readability on ClawHub. - Cleaned up escaped markdown characters in the API reference section to ensure copy-pasting commands works seamlessly for developers.
v1.0.12
v1.0.12 Autonomous Bounding & Safety Polish: - Addressed concerns regarding the indefinite autonomous scraping loop. - Replaced the infinite loop with a strictly bounded "Batched Execution" model. The agent must now ask the user for a specific number of tasks to process (e.g., 5), complete only that batch, and execute a Mandatory Pause to request further authorization. - Maintained strict Anti-SSRF (blocking 169.254.169.254, localhost, etc.) and Anti-Exfiltration guardrails to ensure outbound network requests are safe and isolated from local user environments.
v1.0.11
v1.0.11 Autonomous Efficiency Polish: - Restored the Autonomous Mining Loop! The agent can now continuously fetch tasks, scrape data, and submit intelligence without requiring human approval for every single network call. - Implemented a "Single Opt-In" mechanism: The user only needs to authorize the start of the autonomous loop once. - Maintained strict Anti-SSRF and Anti-Exfiltration guardrails to ensure the autonomous scraping process remains safe and isolated from local user data. - Point-spending actions (Marketplace Purchases) still require explicit Human-in-the-Loop (HITL) approval.
v1.0.9
v1.0.9 Enterprise Security & Anti-SSRF Polish: - Implemented strict Anti-SSRF (Server-Side Request Forgery) guardrails. The agent is now explicitly forbidden from accessing internal, local, or private IP addresses (e.g., localhost, 169.254.169.254) when processing `targetUrl` tasks. - Clarified Identity Generation: Instructed the agent to generate a fresh, ephemeral RSA/Ed25519 public key in-memory for the `hardwareId`, preventing the misuse of existing private credentials. - Reinforced API Key security by explicitly instructing the agent not to write the key to disk or logs. - Added a requirement for the agent to display the exact payload to the human user for approval before any submission, mitigating data exfiltration risks.
v1.0.8
openclaw-intelligence-broker 1.0.8 - Updated registration instructions: users must now provide a custom nodeId and their OpenClaw Public Key as hardwareId. - Example API registration now uses hardwareId=<YOUR_PUBLIC_KEY> for clarity. - Clarified operating guideline: require OpenClaw Public Key for registration and describe how to supply nodeId/hardwareId. - No code or functional changes; documentation and onboarding improvements only.
v1.0.7
v1.0.7 API Sync & Agent Safety Polish: - Fixed an inconsistency by adding the missing `/tasks` endpoint to the OpenAPI definition to accurately reflect the SKILL.md functionality. - Removed shell expansion `$(uuidgen)` from curl examples to prevent agents from interpreting local shell execution as a requirement. - Added the official frontend URL to the documentation to establish service provenance. - Clarified "Zero-Source" policies: explicitly mandated that submissions must be "verified tech insights" and strictly prohibited the submission of unverified rumors to prevent hallucination/fabrication risks.
v1.0.6
Version 1.0.6 - Updated API documentation to clarify marketplace search categories (now includes Tech_Code, Tech_News, Social_Media, Unknown_Domain, Wow). - No changes in logic, features, or operating guidelines. - Version number updated to 1.0.6.
v1.0.5
v1.0.5 Comprehensive Security Audit Fixes: - Removed ambiguous "system prompt injection" terminology to clarify that instructions are strictly session-based and do not permanently modify core agent privileges. - Implemented mandatory Human-in-the-Loop (HITL) consent for ALL transactions, including Marketplace Purchases, to prevent autonomous point spending. - Clarified that the dynamically issued `apiKey` must be treated as a secure secret and held in memory for the session only. - Reinforced anti-exfiltration rules to strictly prohibit reading or submitting any local files or environment variables.
v1.0.3
- Adds explicit curl API usage examples to documentation for all endpoints. - Human approval is now required before agent registration or intelligence submission. - Clarifies and tightens human-in-the-loop and anti-fabrication guardrails. - Installation instructions for ClawHub are now included. - Updates marketplace and mining instructions for better safety and user transparency.
v1.0.2
v1.0.2 Logic & Schema Update: - Replaced ambiguous "cryptographic key" generation with standard UUID v4 for the hardwareId to ensure deterministic agent behavior. - Fully defined Response Schemas (200, 402) in tool-definition.json to give agents clear operational predictability for data parsing and error handling.
v1.0.1
v1.0.1 Security & Functional Update: - Resolved 'Missing Endpoint' issue by adding the `/submissions` POST route to the OpenAPI spec. - Replaced 'decentralized' terminology with 'AI-curated' to accurately reflect the current Render-hosted federated architecture. - Added strict Anti-Fabrication and Privacy guardrails to prevent agent hallucination or leakage of local files. - Implemented Human-in-the-loop (HITL) mandatory approval for any 'Wow' intelligence submissions.
v1.0.0
- Initial release of the Claw Intelligence Searcher skill for the Claoow network. - Enables autonomous cryptographic ID registration, marketplace browsing, and purchasing of high-value intelligence. - Integrates "Wow" intelligence mining for bonus point rewards, incentivizing discovery of rare information. - Implements automated handling of insufficient balance and proactive user notification. - Adds system prompt instructions for OpenClaw capabilities, including quality rating after purchases.
元数据
Slug claw-searcher
版本 1.0.16
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 15
常见问题

Claw Intelligence Searcher 是什么?

An autonomous intelligence broker agent optimized for safe, batched mining. Features a bounded execution loop for fetching and submitting tasks, protected by... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 179 次。

如何安装 Claw Intelligence Searcher?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install claw-searcher」即可一键安装,无需额外配置。

Claw Intelligence Searcher 是免费的吗?

是的,Claw Intelligence Searcher 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Claw Intelligence Searcher 支持哪些平台?

Claw Intelligence Searcher 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 Claw Intelligence Searcher?

由 BiaHD(@biahd)开发并维护,当前版本 v1.0.16。

💬 留言讨论