← 返回 Skills 市场
mirni

Bridge

作者 mirni · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ 安全检测通过
82
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install gh-bridge
功能描述
Agent-to-Human (A2H) verification and escrow platform. Agents request physical-world tasks from humans, define verification criteria (GPS, photos, timestamps...
使用说明 (SKILL.md)

Bridge

Request human help for physical-world tasks. Bridge verifies the work was done before releasing payment.

Start the server

uvicorn bridge.app:app --port 8015

Create a task with verification criteria

curl -s -X POST http://localhost:8015/v1/tasks \
  -H "Content-Type: application/json" \
  -d '{
    "description": "Pick up package at 123 Main St, photograph it",
    "budget_usdc": "25.00",
    "verification_criteria": [
      {"type": "gps_proof", "description": "At pickup location", "params": {"latitude": 37.7749, "longitude": -122.4194, "radius_m": 100}},
      {"type": "photo_proof", "description": "Photo of package", "params": {"min_photos": 1}}
    ]
  }' | jq

Escrow is locked automatically. Fee is 5% of budget.

Submit proof and verify

curl -s -X POST http://localhost:8015/v1/tasks/TASK_ID/verify \
  -H "Content-Type: application/json" \
  -d '{
    "worker_id": "worker-1",
    "proofs": [
      {"type": "gps_proof", "data": {"latitude": 37.7749, "longitude": -122.4194}},
      {"type": "photo_proof", "data": {"photo_hashes": ["sha256:abc123"]}}
    ]
  }' | jq

If ALL criteria pass → escrow released. If any fail → escrow held.

Dispute a task

curl -s -X POST http://localhost:8015/v1/tasks/TASK_ID/dispute \
  -H "Content-Type: application/json" \
  -d '{"reason": "GPS proof appears faked"}' | jq

Freezes escrow. No verification possible while disputed.

Other endpoints

curl -s http://localhost:8015/v1/tasks | jq                    # List all tasks
curl -s http://localhost:8015/v1/tasks?status=posted | jq      # Filter by status
curl -s http://localhost:8015/v1/platforms | jq                 # Available platforms
curl -s http://localhost:8015/v1/stats | jq                    # Platform statistics

Verification types

Type What it proves How
gps_proof Worker was at location Haversine distance \x3C radius
photo_proof Photos submitted Unique hash count >= min
timestamp_proof Done within deadline Elapsed hours \x3C max
signature_proof Cryptographic signature Non-empty signature present

Escrow model

  • Locked at task creation (budget held)
  • Released only if ALL criteria pass
  • Frozen on dispute
  • Refunded if deadline expires with no proof
安全使用建议
This implementation is coherent and appears to be a local/demo A2H escrow/verification service rather than a production-grade payments system. Important cautions: (1) There is no authentication or authorization in the code — if you run uvicorn on a host accessible to others, anyone can list/create/modify tasks. (2) Escrow is simulated in-memory; there is no real payment or blockchain integration. Do not use this as-is for real money without adding secure payment integration, authentication, persistence, TLS, and audit logging. (3) The pip install step pulls packages from PyPI — review package versions and your environment's policy before installing. (4) Proof types (GPS/photos/signatures) can be spoofed; review verification rules and threat model before trusting for high-value tasks. If you plan to deploy beyond local testing, require the maintainer to add authentication, secure storage, real payment connectors, and persistent logs, and audit third-party dependencies.
功能分析
Type: OpenClaw Skill Name: gh-bridge Version: 1.0.0 The gh-bridge skill bundle implements a legitimate Agent-to-Human (A2H) task verification and escrow platform using FastAPI. The code in bridge/app.py and bridge/verifier.py provides structured endpoints and deterministic logic for validating physical-world task proofs (GPS, photos, timestamps) without any evidence of malicious intent, data exfiltration, or unauthorized execution.
能力标签
cryptocan-make-purchases
能力评估
Purpose & Capability
Name/description (Agent-to-Human verification + escrow) match the code and SKILL.md: the service implements task creation, proof submission, verification logic, disputes, and an escrow state machine (in-memory). Required binaries (python) and requested pip packages (fastapi, uvicorn, pydantic) are appropriate for a web API implemented in the provided Python files.
Instruction Scope
SKILL.md instructions are narrowly scoped to starting the local server, creating tasks, submitting proofs, and disputing — exactly what the API implements. The instructions do not ask the agent to read unrelated files, environment variables, or transmit data to unknown external endpoints.
Install Mechanism
Install uses pip to bring in fastapi, uvicorn, and pydantic — a common and expected choice for a Python web API. Pip installs are a moderate-risk install mechanism (packages come from PyPI); verify package sources/versions before installing in sensitive environments.
Credentials
The skill declares no required environment variables or credentials, and the code does not reference environment secrets. There are no requests for unrelated credentials or config paths.
Persistence & Privilege
The skill does not request permanent presence (always=false) and does not modify other skills or system-wide settings. Its state is in-memory only; it does not persist data to disk or external services.
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 gh-bridge
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /gh-bridge 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
v1.0.0
Initial release of gh-bridge: an agent-to-human verification and escrow platform. - Agents can request physical-world tasks with defined verification criteria (GPS, photos, timestamps, signatures). - Automated escrow: funds locked at task creation, released only when all proofs pass. - Built-in dispute mechanism: freezes escrow and verification if challenged. - REST API endpoints to create tasks, submit proofs, dispute, and query platform stats. - Supports multiple verification types for proof: gps, photo, timestamp, and signature.
元数据
Slug gh-bridge
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Bridge 是什么?

Agent-to-Human (A2H) verification and escrow platform. Agents request physical-world tasks from humans, define verification criteria (GPS, photos, timestamps... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 82 次。

如何安装 Bridge?

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

Bridge 是免费的吗?

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

Bridge 支持哪些平台?

Bridge 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 Bridge?

由 mirni(@mirni)开发并维护,当前版本 v1.0.0。

💬 留言讨论