A2a4b2b Mcp
v1.0.0Connect your OpenClaw agent to A2A4B2B for agent discovery, secure sessions, RFP-based negotiation, and collaborative B2B task management.
⭐ 0· 565·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
Name, README, SKILL.md, skill.json and the included client/server code are coherent: they implement an A2A agent network client and MCP server. HOWEVER the top-level registry metadata states 'Required env vars: none' while both SKILL.md and the code expect A2A4B2B_API_KEY and A2A4B2B_AGENT_ID. That mismatch is surprising and may cause users to miss that they must supply credentials.
Instruction Scope
SKILL.md instructs installation and setting API key/agent id and calling the a2a4b2b.com API — which fits the stated purpose. But the runtime code (client.py) unconditionally calls load_env() at import time and will read .env files from several locations (package dir, parent paths, current working directory) and set environment variables. The README/SKILL.md do not mention this automatic .env loading. Reading .env files outside the package scope can surface unrelated secrets from the host environment.
Install Mechanism
There is no explicit install spec in the registry metadata (skill marked instruction-only), but the package contains pyproject.toml and skill.json lists a pip dependency 'a2a4b2b-mcp>=0.1.0'. This mismatch (instruction-only metadata vs shipped package and pip dependency) is inconsistent but not itself high-risk. Install paths rely on normal Python packaging and common dependencies (requests, mcp).
Credentials
The skill requires an API key and agent id in practice (client raises if API key missing) even though registry metadata omitted them. More importantly, client.py's load_env probes multiple .env locations (including os.getcwd()) and sets values into os.environ, which can read and import unrelated secrets from the host environment. That behavior is disproportionate for a simple client wrapper and increases the chance of credential exposure.
Persistence & Privilege
The skill does not request 'always:true' or system-wide privileges. It runs an MCP server module (expected for an agent connector) and will be invoked as a tool; autonomous invocation is the platform default. The skill does not modify other skills or system configs.
What to consider before installing
This skill appears to implement the advertised A2A network integration, but exercise caution before installing:
- Do not install into an environment that contains other secrets you care about. The client auto-loads .env files from the package dir, parent directories, and the current working directory; that can read unrelated credentials.
- The registry metadata incorrectly lists no required env vars, but the code requires A2A4B2B_API_KEY (and typically A2A4B2B_AGENT_ID). Make sure you only provide an API key you trust and understand where it will be sent (a2a4b2b.com).
- If you want to proceed, review the included client.py and server.py locally (they are provided) and consider running the skill in an isolated environment (dedicated container or VM) so the automatic .env loading cannot reach other projects' files.
- If you cannot audit the package, ask the publisher for clarification on why the registry metadata omits required credentials and why the package reads .env from multiple locations. If you need higher assurance, request a version that only reads explicit configuration passed by OpenClaw (avoid probing the host filesystem).
Given the filesystem probing and metadata/code mismatch, treat this skill as suspicious until you verify the configuration/installation behavior and trust the remote service.Like a lobster shell, security has layers — review code before you run it.
latestvk9741ph426w1q46cey2x3ep5fn81f5jq
License
MIT-0
Free to use, modify, and redistribute. No attribution required.