Skip to main content

receipts-guard

Capture and verify all agreements before your agent accepts them. Local risk analysis and evidence capture for autonomous transactions.

Ir a la instalación

Datos de origen

Repositorio
Kernel8901/ai-agent-skills-classification
Última actividad en el origen
4 de abril de 2026 a las 15:26
Idioma detectado de SKILL.md
inglés
Estrellas
4
Forks
0

Opciones de instalación

De forma predeterminada está seleccionado el prompt que primero revisa el origen. Puedes cambiar a un comando directo o descargar una copia local.

Revisa los archivos de origen

Lee SKILL.md y los archivos complementarios que muestra SkillsMP antes de decidir si quieres instalarlo.

Explorador de archivos
3 archivos

Mostrando SKILL.md

SKILL.md
Instrucciones de origen · Vista previa de solo lectura
name
receipts-guard
description
Capture and verify all agreements before your agent accepts them. Local risk analysis and evidence capture for autonomous transactions.
metadata
{"openclaw":{"emoji":"🧾","requires":{"anyBins":"[Truncated]"}}}
# RECEIPTS Guard Protect your agent's autonomy by capturing evidence of every agreement it accepts. When your OpenClaw agent clicks "I agree" or makes a payment, RECEIPTS creates a local, immutable record. **Problem solved:** When disputes happen, you have proof of exactly what was agreed to. **No API. No cloud. Your data stays local.** ## Why You Need This Your OpenClaw agent is: - Accepting Terms of Service without you knowing the full terms - Making purchases with binding agreements - Signing up for services with hidden clauses Every action creates liability. Zero creates evidence. Until now. ## Configuration Configure in **`~/.openclaw/openclaw.json`**: ```json "skills": { "entries": { "receipts-guard": { "enabled": true, "env": { "RECEIPTS_AGENT_ID": "my-openclaw-agent" } } } } ``` | Variable | Description | |----------|-------------| | RECEIPTS_AGENT_ID | Unique identifier for this agent instance (optional) | ## Agent Instructions ### Before Accepting Any Terms of Service 1. **Extract the terms text** from the page before clicking any "I agree", "Accept", or "Continue" button 2. **Run the capture script** to analyze and store the agreement: ```bash node ~/.openclaw/workspace/skills/receipts-guard/capture.js "TERMS_TEXT" "SOURCE_URL" "MERCHANT_NAME" ``` 3. **Check the response**: - If `recommendation: "proceed"` - Safe to accept - If `recommendation: "require_approval"` - Ask user before accepting - If `recommendation: "block"` - Do NOT accept, inform user of risk flags 4. **Store the receipt ID** for future reference ### Before Making Any Payment 1. **Capture the terms** associated with the purchase 2. **Include transaction details** when available: ```bash node ~/.openclaw/workspace/skills/receipts-guard/capture.js "CHECKOUT_TERMS" "https://merchant.com/checkout" "Merchant Name" ``` 3. **Only proceed with payment** if capture succeeds and recommendation allows ### Security Rules - **Never skip capture** for agreements or payments - this protects the user - **Never ignore block recommendations** - these indicate serious risk flags - **Always store receipt IDs** in transaction logs for future disputes ## Usage Examples ### Capture Terms of Service ```bash node ~/.openclaw/workspace/skills/receipts-guard/capture.js \ "By using this service you agree to binding arbitration..." \ "https://example.com/terms" \ "Example Corp" ``` ### Response Format ```json { "captureId": "local_6c6870d3e01cd265", "recommendation": "proceed", "trustScore": 72, "riskFlags": [], "summary": "Standard terms with 30-day refund policy", "documentHash": "6c6870d3e01cd265..." } ``` ### Risk Flag Response ```json { "captureId": "local_abc123def456", "recommendation": "block", "trustScore": 40, "riskFlags": [ "Binding arbitration clause", "Class action waiver", "No refund policy" ], "summary": "Terms contain multiple restrictive clauses" } ``` ## What Gets Captured - Full document text (SHA-256 hashed for immutability) - Source URL and timestamp - Merchant/service name - Risk analysis results - Trust score All data stays local on your machine. No external API calls. ## Risk Flags Detected The local analyzer flags: - Binding arbitration clauses - Class action waivers - Rights waivers - No refund policies - Auto-renewal clauses - Data selling/sharing clauses - Limited liability clauses - Indemnification clauses ## Links - SDK: `npm install @lazaruseth/agreement-guard` - GitHub: https://github.com/lazaruseth/receipts-mvp ## Troubleshooting - **Capture failed**: Ensure Node.js is installed and the script path is correct - **No terms found**: Ensure you're extracting the full terms text before capture
Ver en GitHub