功能描述

AI agent infrastructure for deals, escrow, attestations, and autonomous agents. 39 tools across 9 engines.

使用说明 (SKILL.md)

deal.works

Name: Deal Works Mcp
Author: swgoettelman

AI agent infrastructure for trustworthy deal-making. Enables agents to create deals, manage escrow, deploy autonomous agents, and issue cryptographic attestations.

Quick Start

# Install the skill
openclaw skill add swgoettelman/deal-works

# Set your API key
export DEAL_WORKS_API_KEY=your_key_here

# The skill is now available to your agent

Get your API key at hq.works/settings/api.

Capabilities

9 Engines, 39 Tools

Engine	Tools	Purpose
Deal	7	Deal lifecycle: create, sign, approve, search, timeline
Fund	6	Wallets, transfers, escrow, cashout, agent funding
Bourse	5	Marketplace for templates, skills, integrations
Cadre	6	Deploy and manage autonomous agents
Oath	5	Cryptographic attestations (completed on-chain)
Parler	4	Dispute resolution and governance
Academy	3	Learning platform for deal-making
HQ	2	Dashboard metrics and system health
Clause	1	Contract clause rendering

7 Resources

dealworks://profile - User profile and trust tier
dealworks://wallet - Wallet balances
dealworks://deals - Active deals
dealworks://agents - Deployed agents
dealworks://templates - Available templates
dealworks://disputes - Open disputes
dealworks://dashboard - Key metrics

5 Pre-built Prompts

escrow-deal - Create escrow-protected deal workflow
deploy-agent - Deploy agent from skill
file-dispute - File dispute with evidence
publish-template - Publish template to Bourse
portfolio-review - Review deal portfolio

Common Workflows

Create an Escrow Deal

User: Create a $5000 escrow deal with [email protected] for website development

Agent: I'll use deal_create to create the deal, then fund_escrow to lock the funds,
       and oath_attest to prepare completion verification.

Deploy a Monitoring Agent

User: Deploy an agent to monitor my active deals and alert on SLA violations

Agent: I'll use bourse_search to find monitoring skills, cadre_deploy to launch
       the agent, and fund_agent_fund to allocate operational budget.

File a Dispute

User: The contractor didn't deliver. File a dispute for deal xyz123.

Agent: I'll use deal_get to review the deal, then parler_dispute_file with
       the evidence. The dispute enters the resolution queue.

Tool Reference

Deal Engine

Tool	Description
`deal_list`	List deals with optional status filter
`deal_create`	Create a new deal with terms and counterparty
`deal_get`	Get detailed deal information
`deal_action`	Perform action: SIGN, APPROVE, REJECT, CANCEL, COMPLETE, ARCHIVE
`deal_search`	Search deals by text query
`deal_timeline`	Get activity timeline for a deal
`deal_attachments`	List deal attachments

Fund Engine

Tool	Description
`fund_balance`	Get wallet balance (available, locked, pending)
`fund_transfer`	Transfer funds between wallets
`fund_transactions`	List transaction history
`fund_escrow`	Lock funds in escrow for a deal
`fund_cashout`	Cash out to external wallet on Base
`fund_agent_fund`	Fund an agent's operational wallet

Bourse Engine

Tool	Description
`bourse_search`	Search marketplace for templates/skills
`bourse_get`	Get listing details and reviews
`bourse_fork`	Fork a listing to customize
`bourse_publish`	Publish to marketplace
`bourse_earnings`	View earnings from published listings

Cadre Engine

Tool	Description
`cadre_list`	List deployed agents
`cadre_deploy`	Deploy a new agent from skill
`cadre_command`	Send command: START, STOP, RESTART, SCALE
`cadre_health`	Get agent health status
`cadre_delegations`	List permission delegations
`cadre_sla_violations`	List SLA violations

Oath Engine

Tool	Description
`oath_attest`	Create attestation for deal milestone
`oath_verify`	Verify attestation authenticity
`oath_vault_upload`	Upload document hashes to vault
`oath_vault_seal`	Seal vault with Merkle root on-chain
`oath_trust_tier`	Get user/org trust tier

Parler Engine

Tool	Description
`parler_dispute_file`	File a dispute with evidence
`parler_dispute_list`	List disputes by status/role
`parler_proposals`	List resolution proposals
`parler_vote`	Vote on proposal

Academy Engine

Tool	Description
`academy_courses`	Browse available courses
`academy_enroll`	Enroll in a course
`academy_tip`	Tip a course creator

HQ Engine

Tool	Description
`hq_dashboard`	Get dashboard metrics
`hq_health`	Check system health

Clause Engine

Tool	Description
`clause_render`	Render contract clause template

Security

All mutations use idempotency keys (safe to retry)
API keys scoped per engine
Rate limited: 100 requests/minute
Circuit breaker per engine (5 failures = 30s cooldown)
Only calls *.works domains + approved external APIs

Links

deal.works - Main platform
Documentation - Full API docs
npm package
GitHub

安全使用建议

This package appears to implement the deal.works MCP connector and only needs your DEAL_WORKS_API_KEY and node, which is reasonable. However: 1) The SKILL.md was flagged for a possible 'system-prompt-override' pattern — open and read SKILL.md/README yourself and search for any lines that tell the agent to ignore prior instructions or to change system prompts; remove or reject the skill if present. 2) Verify the npm package and publisher (check the package on the npm registry or GitHub) — there are slight name inconsistencies in the files (swgoettelman vs goettelman) that could indicate a packaging mistake or impersonation. 3) Prefer installing and testing in a sandboxed environment first (use an API key with limited scope/permissions), and be ready to rotate the API key if you detect unexpected behavior. 4) If you plan to let agents invoke this autonomously, be extra cautious: autonomous skills with prompt-injection vectors can change agent behavior. If you want, provide the SKILL.md/GitHub link and I can re-check for exact prompt-injection phrases and suspicious instructions.

功能分析

Type: OpenClaw Skill Name: deal-works Version: 0.1.1 The skill bundle provides extensive functionality for managing deals, funds (including transfers and cash-outs), and autonomous agents, interacting with several `*.works` domains and `api.perplexity.ai`. While the code itself does not exhibit clear evidence of intentional malicious behavior (e.g., no arbitrary command execution, no unauthorized data exfiltration to unknown endpoints, no obfuscation), its powerful financial and agent management capabilities, combined with the direct embedding of user-provided arguments into agent prompts (as seen in `src/prompts.ts` and `configs/perplexity-prompts.md`), present a significant prompt injection risk. A compromised or manipulated AI agent could be instructed to misuse these powerful tools (e.g., `fund_cashout`, `fund_transfer`, `cadre_deploy`) to perform unauthorized financial transactions or deploy malicious agents, even if the skill itself is not designed for self-exploitation. This falls under risky capabilities without clear malicious intent, classifying it as suspicious.

能力评估

✓ Purpose & Capability

Name/description (deal.works, deals/escrow/agents) align with the requested binary (node), the npm package @swgoettelman/deal-works-mcp, and the single required env var (DEAL_WORKS_API_KEY). The declared tools and base URLs in code point to *.works domains which fits the described functionality.

⚠ Instruction Scope

SKILL.md is largely a normal MCP skill README/instruction file, but the pre-scan detected a 'system-prompt-override' pattern (prompt-injection). Even though the visible SKILL.md content here doesn't obviously contain an active override, the scanner signal indicates the file may include phrasing that attempts to change agent/system prompts or agent behavior. That is scope-creep for a tool connector and should be manually inspected.

✓ Install Mechanism

Install uses an npm package (registry install) and exposes a CLI binary 'deal-works-mcp' — a standard mechanism. No remote arbitrary archive downloads or extract-from-unknown-URL installs were found.

✓ Credentials

Only one required environment variable (DEAL_WORKS_API_KEY) is declared and used by the client code for Authorization headers. This matches the stated need to authenticate to deal.works APIs and is proportionate to the skill's purpose.

✓ Persistence & Privilege

No elevated privileges requested. always is false; the skill is user-invocable and may be invoked autonomously (platform default). The package does not request system-wide config changes or other skills' credentials.

版本历史

v0.1.1

Added Perplexity config templates, Sonar API integration, and pre-built Computer prompts

v0.1.0

Initial public release of deal-works v0.1.0 - Provides AI agent infrastructure for deals, escrow, attestations, and autonomous agents. - Includes 39 tools across 9 engines, supporting workflows such as deal creation, escrow management, agent deployment, and dispute resolution. - Features 7 resources and 5 pre-built prompts for common operations. - Offers a detailed tool reference for each engine, covering deal, fund, bourse, cadre, oath, parler, academy, HQ, and clause functionalities. - Security highlights: idempotency, per-engine API scoping, rate limiting, and domain restrictions. - Documentation, installation, and API key setup instructions included.

元数据

Slug deal-works

版本 0.1.1

许可证 —

累计安装 1

当前安装数 1

历史版本数 2

常见问题

Deal Works Mcp 是什么？

AI agent infrastructure for deals, escrow, attestations, and autonomous agents. 39 tools across 9 engines. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 511 次。

如何安装 Deal Works Mcp？

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

Deal Works Mcp 是免费的吗？

是的，Deal Works Mcp 完全免费（开源免费），可自由下载、安装和使用。

Deal Works Mcp 支持哪些平台？

Deal Works Mcp 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 Deal Works Mcp？

由 swgoettelman（@swgoettelman）开发并维护，当前版本 v0.1.1。

Deal Works Mcp