⚡ Start in 60 seconds
1 · Install the skills (needs Node 18+ and git on PATH):
npx skills add lnvestor/twitr-skills
2 · Get a wallet. Any x402/MPP wallet works — each one's own documented command:
First time with agent wallets? → Wallet setup has step-by-step guides (AgentCash's onboarding bonus can even fund you up to $25 for free).
3 · Ask your agent something. The first call returns HTTP 402 with the exact price — that's the quote, not an error. Your wallet pays and retries automatically. $5 covers thousands of reads.
💡 What can my agent do?
Copy-paste any of these once the skills are installed:
"What is @vercel posting about this week?"
"Search X for what people are saying about my product — top 50 by likes, English only."
"Watch @competitor and tell me the moment they announce something." — a real-time monitor; events reach your agent by free polling or signed webhook
"Draft 3 tweet variants announcing our launch in my voice — I'll pick one, then post it."
"Export everyone who replied to this tweet as a dataset."
"Run my X presence: watch these 3 topics, queue drafts every morning, replies wait for my approval."
📦 The skills
Five focused skills with explicit trigger boundaries, so they don't fire over each other.
| Skill | What it does |
|---|
| ⭐ | x-presence | Unattended presence — watch topics, draft in your voice, publish on a schedule, queue replies for approval. Start here — it drives the others internally. |
| 🔎 | x-twitter-data | Live reads — tweets, profiles, search, timelines, lists, communities, trends. |
| 🔔 | x-twitter-monitor | Real-time monitors, delivered by free polling or HMAC-signed webhooks. |
| ✍️ | x-twitter-publish | Draft with AI, then post, reply, or quote through a connected account. |
| 📦 | x-twitter-research | Bulk exports — followers, repliers, reposters — as downloadable datasets. |
Install one instead of all five with --skill <name>.
Scheduling x-presence? See scheduling.md — verified commands for Claude Code routines, Hermes cron, and OpenClaw cron.
👛 Wallet setup
AgentCash — one command, ~2 minutes, and it may pay you up to $25.
1. Onboard (creates a local wallet, installs the AgentCash skill, and wires up MCP for your agent):
npx agentcash@latest onboard
Visiting agentcash.dev/onboard first and connecting GitHub/X/LinkedIn
gets you a claim code worth up to $25 in free USDC — often enough to run these skills for weeks
without depositing anything.
2. Fund it (skip if the bonus covered you):
Guided flow with card on-ramp and bridging. No minimum.
3. Check it works — ask your agent "What is my AgentCash balance?" If it answers with a number,
you're done.
Claude Code users can alternatively install just the MCP server:
claude mcp add agentcash --scope user -- npx -y agentcash@latest
Coinbase Agentic Wallet (awal) — email OTP login, fund with Apple Pay/card.
1. Sign in (a 6-digit code lands in your email; verify it):
npx awal@latest auth login
npx awal@latest auth verify <flowId> <otp>
2. Fund it — npx awal@latest show opens the Coinbase Onramp UI (Apple Pay, card, bank).
3. Done. Your agent pays twitr.sh over x402 on Base; to pay an endpoint by hand:
npx awal@latest x402 pay <url>.
pay.sh — Solana Foundation's payment layer; wraps your agent CLI.
1. Set up the wallet, then wrap the tool you use:
pay detects the 402/MPP challenge, asks your local wallet to sign, and retries with
payment proof — USDC on Solana. twitr.sh accepts Solana x402 natively.
Your client's per-call cap is a hard spend limit — no call can exceed it, and failed calls are never charged.
💸 What it costs
| Action | Price |
|---|
| Reading events from a monitor | free |
| Drafting a post | ~$0.001 |
| A read or search | ~$0.0012 / item |
| A post or reply | $0.036 |
| A monitor | ~$0.025 / hour |
A typical week running x-presence — two monitors, a dozen replies, a few posts — is about
$9. Every 402 quotes the exact price before your wallet signs anything.
🚫 What these skills will not do
No automated following, unfollowing, or liking, and no keyword-triggered replies.
X prohibits these by name, and its published ranking weights show follows aren't scored at all
while a single report scores −369.0 — so the banned tactics cost reach as well as risking
the account. The prohibited actions are omitted from the skills rather than merely discouraged.
Reasoning and sources: ranking-signals.md ·
x-rules.md
🔗 Links
License
MIT — see LICENSE.
Independent service. Not affiliated with X Corp. "X" and "Twitter" are trademarks of X Corp.
Enjoying twitr-skills? A ⭐ on this repo helps other agents find it.
