← 返回 Skills 市场

a2a-Market-ACP-Lite-Negotiation

Name: a2a-Market-ACP-Lite-Negotiation
Author: luoqianchenguni-max

作者 luoqianchenguni-max · GitHub ↗ · v0.2.8 · MIT-0

cross-platform ⚠ suspicious

296

总下载

当前安装

版本数

在 OpenClaw 中安装

/install a2a-market-acp-lite-negotiation

功能描述

Gateway-only ACP negotiation skill with optional OpenClaw model-driven turn decisions.

使用说明 (SKILL.md)

a2a-Market ACP Lite Negotiation

This skill is now gateway-only. Single-turn local decision mode has been removed.

Gateway Loop Mode (only mode)

Start one participant in one terminal:

node src/cli/index.js --role buyer --agent-id buyer-openclaw --gateway http://127.0.0.1:3085

Flow:

POST /agents/register
Loop GET /agents/pull
POST /agents/respond
For NEGOTIATION_TURN, decide by selected engine (rule or openclaw)

Core flags

--role buyer|seller
--gateway (default http://127.0.0.1:3085)
--agent-id
--decision-engine rule|openclaw (default rule)
--auth-token (default market-auth-token)
--pull-timeout-ms (default 25000)
--retry-delay-ms (default 1000)
--max-polls (0 means infinite)
--verbose true|false

OpenClaw engine flags

--provider-env (default OPENAI_API_KEY)
--api-key (optional direct key)
--allow-no-key true|false (default false)
--thinking (default low)
--timeout seconds (default 60)
--openclaw-extra-prompt (optional)

Auto-start session flags

--start-session true|false
--counterparty-agent-id (required with --start-session true)
--list-amount-minor-units (default 9000)
--currency (default USD)
--max-rounds (default 5)
--product (optional)
--goal (optional)
--floor-minor-units (optional)
--ceiling-minor-units (optional)
--session-id (default nego_\x3Ctimestamp>)
--wait-counterparty-ms (default 15000)
--stop-on-session-end true|false

Single terminal: run skill + model + auto-kickoff

node src/cli/index.js \
  --role buyer \
  --agent-id buyer-openclaw \
  --gateway http://127.0.0.1:3085 \
  --decision-engine openclaw \
  --start-session true \
  --counterparty-agent-id seller-openclaw \
  --list-amount-minor-units 9000 \
  --currency USD \
  --max-rounds 5 \
  --wait-counterparty-ms 30000

Executable Entrypoint

node src/cli/index.js
```\r

安全使用建议

This skill is generally what it says: a gateway-focused negotiation CLI that can optionally use an external LLM provider to decide turns. However, note these practical and privacy points before installing or running it: - The package metadata does not declare Node.js, but the script is a Node CLI. You must run it with node (e.g., node src/cli/index.js). - If you enable the openclaw decision engine, the tool will look for LLM API keys in environment variables (OPENAI_API_KEY, ANTHROPIC_API_KEY, GOOGLE_API_KEY, etc.) or accept a key via --api-key. Those keys (and the negotiation context) will be sent to whichever provider you configure — only supply keys you trust the provider with. - Avoid passing API keys on the command line if possible; command-line arguments can be visible to other local users/processes. Prefer environment variables set in a secure way. - The skill communicates with a gateway URL (default http://127.0.0.1:3085). Ensure the gateway endpoint is trusted; message payloads (session, offers, utterances) will be transmitted to it. - The bundle includes source code; if you are concerned about hidden behavior, inspect the remainder of src/cli/index.js (the file was truncated in the provided extract) or run the tool in a sandboxed environment and observe outbound network connections. If you plan to use this skill, run it in an isolated environment (or audit the full source) and only provide LLM API keys and gateway endpoints you trust.

功能分析

Type: OpenClaw Skill Name: a2a-market-acp-lite-negotiation Version: 0.2.8 The skill bundle provides a CLI tool for automated Agent-to-Agent (A2A) price negotiations. It supports rule-based logic or LLM-driven decisions by spawning the 'openclaw' binary (src/cli/index.js) and communicating with a negotiation gateway via HTTP. While the script handles sensitive API keys and executes local commands, these behaviors are documented and necessary for its stated purpose; no evidence of data exfiltration, malicious prompt injection, or intentional backdoors was found.

能力评估

ℹ Purpose & Capability

The SKILL.md and code implement a gateway-only negotiation CLI (expected network calls to the gateway and optional model-driven decisions). However, the registry metadata lists no required binaries or env vars while the README and code clearly expect to be run with node and optionally use provider API keys (OPENAI_API_KEY etc.). The missing declaration of Node as a required binary and omission of optional provider env variables is an incoherence.

✓ Instruction Scope

The SKILL.md limits behavior to registering with a gateway, polling for turns, and posting responses; the code shown implements JSON parsing, decision normalization, and optional model-driven parsing. The runtime also reads stdin, command-line flags, and environment variables to build context — all of which are consistent with a CLI agent that interfaces with a gateway and optionally an LLM provider.

✓ Install Mechanism

There is no install spec (instruction-only deployment), which is low-risk. The bundle does include a Node.js script (src/cli/index.js) that must be executed with node; packaging omitted an explicit 'requires node' declaration. No remote download/install steps are present.

ℹ Credentials

The registry lists no required env vars but the code explicitly inspects many common LLM provider env names (OPENAI_API_KEY, ANTHROPIC_API_KEY, GOOGLE_API_KEY, etc.) and the SKILL.md documents provider flags. Requesting or using an LLM API key is proportionate to the 'openclaw' decision engine, but the metadata omission means users may not realize they need to provide sensitive credentials. Also passing an API key via the --api-key CLI flag will expose it via process arguments (visible to local users).

✓ Persistence & Privilege

The skill is not marked always:true and does not claim to modify other skills or global agent configuration. It runs as a user-invoked CLI and requires no permanent platform presence.

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 a2a-market-acp-lite-negotiation
安装完成后，直接呼叫该 Skill 的名称或使用 /a2a-market-acp-lite-negotiation 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

v0.2.8

gateway-only mode: remove standalone single-turn decision flow; enforce register/pull/respond session loop

v0.2.7

add decision-engine=openclaw for model-driven turn decisions in run-skill gateway loop

v0.2.6

add in-skill session start via /sessions/negotiate for single-terminal run-skill flow

v0.2.5

add gateway agent loop mode for run-skill register/pull/respond while preserving turn-decision contract

v0.2.4

switch to turn-decision-v1 single-turn decision contract and executable local strategy CLI

v0.2.3

add executable skill entrypoint src/cli/index.js and backend negotiation runner

v0.2.2

fix OpenClaw execution: remove local src dependency, run one-call backend negotiation via gateway HTTP

v0.2.1

enable one-call backend negotiation flow with role/product/price params and trace output

v0.2.0

sync runtime implementation and validation coverage

v0.1.0

initial scaffold for early registration

元数据

Slug a2a-market-acp-lite-negotiation

版本 0.2.8

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 10

常见问题