Agent wallet, identity, and paid tools in one package. Register an agent, fund it via Stripe or USDC, then use the balance for web search, AI image generation, AI video generation, AI music creation, X/Twitter search, email send/receive, SMS and voice calls, contacts management, and 100+ LLM models. The funding and identity layer for autonomous agents that need to spend money, send messages, make phone calls, or call paid APIs.
Detected risks:
ATXP is an agent-infrastructure skill that gives an autonomous agent a funded identity and access to paid tools through a single CLI (invoked via npx). It bundles a wallet, an identity, and an email/phone address so an agent can register, fund itself via Stripe or USDC, and then spend that balance on paid capabilities: web search, AI image/video/music generation, X/Twitter search, sending and receiving email, SMS and voice calls, contacts management, and access to many LLM models. It targets the concrete gap of letting agents transact and communicate autonomously rather than being read-only.
Operationally, the skill runs `npx atxp@latest` subcommands, storing an authentication token (ATXP_CONNECTION) in a plain-text ~/.atxp/config file and local contacts in ~/.atxp/contacts.json, and communicating with *.atxp.ai over HTTPS. Notably, a large part of the SKILL.md is a detailed security model: it flags which commands return untrusted external content (search results, tweets, inbound email/SMS, attachments, call transcripts) that may carry prompt injection, and lays out rules against executing untrusted commands, leaking the ATXP_CONNECTION credential, exfiltrating config or secrets via email/SMS, and performing financial actions in response to external instructions. Paid commands are explicitly labeled, with guidance to verify intent and check balance before spending.
Target users are developers building autonomous or semi-autonomous agents that must spend money, send messages, place calls, or call paid APIs. The skill is powerful and inherently dual-use: it can move real funds and send outbound communications, and it downloads and runs code via npx. Those genuinely elevated capabilities lower its score somewhat, but they are paired with unusually strong, security-conscious guardrails rather than any malicious or exfiltration-oriented design.
Register an identity, fund a wallet via Stripe or USDC, and spend the balance on web search, AI image/video/music generation, X/Twitter search, email and SMS, voice calls, contacts management, and 100+ LLM models.
Node.js >=18 and npx, plus HTTPS network access to *.atxp.ai. The authentication token ATXP_CONNECTION is stored in a plain-text ~/.atxp/config file and should be treated as a sensitive secret set via environment variable, never as a CLI argument.
It explicitly marks commands that return external content (search, tweets, inbound email/SMS, attachments, call transcripts) as untrusted, and instructs the agent never to execute verbatim commands from them, to wrap them in boundary markers, and to ignore embedded directives.
It controls a funded wallet and can send email/SMS and place calls, so misuse could spend money or transmit data. The skill mitigates this by labeling paid commands, forbidding financial or send actions triggered by external content, and barring transmission of credentials or config.
Yes—it executes npx atxp@latest, which downloads and runs the atxp package from the npm registry each invocation, so it depends on the integrity of that published package.
Quick Setup:
.claude/skills/Repository
atxp-dev/cli