CreditClaw Amazon

SuspiciousAudited by ClawScan on May 10, 2026.

Overview

This skill is a real checkout/payment integration, but it asks the agent to handle card, wallet, account, and public payment actions beyond the advertised Amazon-checkout scope.

Review this as a broad financial/payment automation skill, not just an Amazon checkout helper. Before installing, confirm you trust CreditClaw, use strict owner approvals and spending limits, avoid persistent memory for card data, and do not let the agent follow remote vendor instructions or send invoices/payment links without explicit confirmation.

Findings (6)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

Installing this may give an agent authority to operate a broader financial/payment account than a user would expect from an Amazon checkout skill.

Why it was flagged

The skill presents as an Amazon checkout skill but documents much broader financial authority, including card spending, wallet payments, and storefront management.

Skill content
name: amazon-checkout ... description: Amazon Checkout ... CreditClaw.com is a financial enablement platform ... Accept card details ... make purchases ... A stablecoin wallet ... Storefronts and product management
Recommendation

Only install if you intend to use the full CreditClaw financial platform. Keep approval mode set to ask for every purchase, set strict spending limits, and ensure the registry description and credential declarations accurately reflect the real scope.

What this means

A remote vendor instruction could change after installation and influence a high-impact checkout flow without being statically reviewed here.

Why it was flagged

The skill tells the agent to fetch and follow remote, mutable checkout instructions that are not included in the reviewed files.

Skill content
If a vendor skill exists → use it ... Get a Vendor Skill ... Returns the vendor's complete checkout instructions as Markdown.
Recommendation

Treat remotely fetched vendor instructions as untrusted until reviewed, require explicit user confirmation before following them, and prefer pinned or locally bundled versions for payment workflows.

What this means

An agent with the API key could create public payment artifacts or send invoices, creating external business or reputational effects beyond simple shopping.

Why it was flagged

The bundled docs expose public payment-page and email-invoice actions that are not clearly part of an Amazon checkout task.

Skill content
Create public checkout pages where anyone can pay you ... Send Invoices ... POST /bot/invoices/[id]/send ... Sends the invoice to the recipient via email
Recommendation

Separate selling/invoicing capabilities into a distinct skill or require explicit, per-action user approval before creating checkout pages, payment links, or sending invoices.

What this means

If approvals or spending limits are too loose, the agent could complete real purchases with the owner's funds.

Why it was flagged

The core checkout flow intentionally lets the agent drive browser/API actions to complete purchases. This is purpose-aligned, but financially sensitive.

Skill content
Once approved → call POST /bot/rail5/key ... Decrypt card details ... Fill shipping/billing, then card fields ... Submit and capture confirmation
Recommendation

Keep owner approval enabled, verify merchant, item, and amount before checkout, and stop on CAPTCHA, OTP, 3DS, or unexpected order-review screens.

What this means

Card data could be exposed if logs, transcripts, memory, or screenshots capture it during checkout.

Why it was flagged

The skill explicitly handles decrypted card data in agent/browser context, while also documenting appropriate ephemeral handling.

Skill content
Never store, log, or persist decrypted card data. It exists only in memory for the duration of a single checkout. Discard it immediately after.
Recommendation

Use ephemeral checkout contexts or sub-agents, disable persistent memory/logging for card entry, and verify that decrypted card values are never saved in chat history or files.

What this means

A misconfigured or untrusted callback URL could receive sensitive sales, payment, or balance information.

Why it was flagged

The skill documents webhook delivery of financial and buyer information to a configured callback URL.

Skill content
CreditClaw fires webhooks to your callback_url ... wallet.sale.completed ... buyer_email ... new_balance_usd
Recommendation

Only set callback URLs you control, store webhook secrets securely, and verify webhook signatures before acting on events.