ClawTrust
1.8k Downloads
Overview
Enable AI agents to autonomously search for, budget-check, and purchase Amazon products using a credit-backed mUSD wallet on Base mainnet, with optional proactive (auto) purchasing and support workflows.
Key Advantages
1.End-to-end purchase flow coverage: from agent registration and credential management through catalog search, quoting, payment signing (x402), and order placement.
2.Built-in credit and budget control model, including per-order and daily caps, auto top-ups, and explicit budget_check() before purchases.
3.Optional proactive purchasing with a clearly defined confidence model, category restrictions, price caps, and system blocklists for high-risk goods.
4.Robust error-handling design (rate limits, quote expiry auto-retry, insufficient credit, invalid shipping, suspension, and token pauses) with clear user-facing recovery guidance.
5.Separation of concerns for sensitive data via a web portal (cards, full account management) and local credentials file with appropriate permissions and paths per OS, plus idempotency guidance for all
Use Cases
- Household restocking assistant that automatically orders low-risk essentials such as toiletries, cleaning supplies, and basic kitchen items within a configured budget.
- Office or team operations bot that monitors shortages (paper, cables, adapters, basic peripherals) and autonomously replenishes them under spending caps.
- Shopping copilot for users who want the agent to search Amazon, propose items with final prices, and then execute the purchase once confirmed.
- Developer or research automation where an AI workflow can trigger hardware or supply orders as part of a broader pipeline, while respecting credit limits.
- Customer-support-style agent that can open support tickets for returns, lost packages, or billing questions linked to prior orders.
Evaluation Scores
7.7
/ 10
Reliability
7.5
Functionality
8.5
Usability
7.8
Safety
6.8
Performance
8.0
Compatibility
8.2
Based on 1 evaluation · Latest: 3/19/2026
Download Trend
Loading...
Evaluation History (1)
7.7/103/19/2026▼
OS: win32-x64LLM: anthropic/claude-sonnet-4.6
**Judgement:** Moltpho is a sophisticated, well-specified Amazon shopping skill suitable for agents that need real monetary autonomy under a credit model. It provides a full purchase lifecycle (registration, search, quoting, payment, orders, support) and integrates cleanly with OpenClaw-style agents via HTTP APIs and a local credentials file.
**Key strengths:**
- Comprehensive functional coverage: bootstrap, shipping profile upsert, catalog search with constraints, budget checking, multi-step x402 payment flow, and support ticketing.
- Strong operational detail: clear rate limits, quote TTL behavior with auto-retry and price-tolerance rules, explicit agent states (UNCLAIMED/CLAIMED/DEGRADED/SUSPENDED), and thorough error mapping with prescribed recovery steps.
- Sensible safety scaffolding for autonomy: proactive purchasing restricted to low-risk categories, price caps (min(per_order_cap, $75)), confidence thresholds (≥0.8) and explicit handling of budget-related language to down-weight purchase decisions.
**Primary risks and concerns:**
- **Financial and autonomy risk:** The skill enables real-money purchases via mUSD on Base. Proactive purchasing is ON by default, so misclassification of user intent or mis-set credit limits could still cause unwanted orders, despite the confidence model and caps.
- **Scope and geographic limits:** Shipping is US-only in v1, which may frustrate non-US users and must be checked before promising fulfillment.
- **Dependency stack complexity:** Reliability depends on multiple external systems (Moltpho API, Amazon catalog, Base mainnet, Stripe, portal). Failures are anticipated in the design but still create user-friction and potential partial states (e.g., DEGRADED or TOKEN_PAUSED).
**Recommended scenarios:**
- Agents that act as **controlled auto-replenishers** for low-risk, recurring needs (household or office supplies) where the owner is comfortable granting limited spending autonomy.
- **Shopping copilots** where the agent primarily searches and proposes purchases, using explicit human confirmation before calling `purchase()`.
- **Operational or dev automation** where predictable small hardware/supply needs must be ordered by workflows, and administrators can manage credit limits via the portal.
In more safety-sensitive deployments (e.g., broad consumer assistants or users with tight financial constraints), this skill should be used with proactive purchasing disabled or heavily constrained, and with clear prompts around confirming every purchase and respecting budget signals.
Comments (0)
No comments yet. Be the first!