メニュー
Aggregate, score, and prioritize social mentions across X/Twitter. Outputs a ranked inbox with engagement scores and draft context for efficient reply batching.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Encode EVM function calls and constructor args from signature + values. Zero deps.
Find Uniswap V2/V3 liquidity pools for any token on Base or Ethereum — reserves, pricing, TVL
Inspect Uniswap Permit2 sub-approvals for any wallet on Base or Ethereum
Resolve EVM function selectors (4-byte) and event topic hashes to human-readable signatures. Queries openchain.xyz + 4byte.directory with a built-in fallback DB of 50+ common signatures. Batch mode, bytecode scanning, stdin pipe. Zero dependencies.
Resolve Basenames (.base.eth) to addresses and back on Base
Turn raw EVM revert data into a readable error. Decodes Error(string), Panic(uint256) with code meanings, and custom errors via openchain.xyz selector lookup. Zero dependencies. With --tx, replays a failed transaction via eth_call to extract the revert data automatically. Use when an agent's tx reverted and the only signal is an opaque 0x… blob.
| name | social-inbox |
| description | Aggregate, score, and prioritize social mentions across X/Twitter. Outputs a ranked inbox with engagement scores and draft context for efficient reply batching. |
Fetches X/Twitter mentions and brand searches, deduplicates against already-replied tweets, scores by engagement potential, and outputs a prioritized inbox.
# Full scan: mentions + brand search
node ~/Github/axiom-public/agent-skills/skills/social-inbox/social-inbox.mjs
# Mentions only
node ~/Github/axiom-public/agent-skills/skills/social-inbox/social-inbox.mjs --mentions-only
# Output as JSON (for piping)
node ~/Github/axiom-public/agent-skills/skills/social-inbox/social-inbox.mjs --json
# Custom search terms
node ~/Github/axiom-public/agent-skills/skills/social-inbox/social-inbox.mjs --search "AppFactory,axiom ventures"
Saves to ~/clawd/data/social-inbox.json with structure:
{
"lastScan": "2026-02-08T...",
"items": [
{
"id": "tweet_id",
"author": "@handle",
"text": "...",
"url": "https://x.com/...",
"score": 8.5,
"scoreBreakdown": { "relevance": 3, "question": 2, "directMention": 2, "recency": 1.5 },
"status": "pending",
"source": "mentions"
}
]
}
| Signal | Points |
|---|---|
| Direct @mention | +2 |
| Question mark (asking us something) | +2 |
| Keywords: fund, invest, agent, build, LP, token | +1 each (max 3) |
| Recency (<1h: +2, <6h: +1.5, <24h: +1) | up to +2 |
| Not a bot/spam pattern | +1 (default, -1 if detected) |
Items scoring <3 are filtered out. Already-replied tweets are excluded.
Add to OpenClaw cron for periodic scanning:
{
"kind": "agentTurn",
"message": "Run social inbox scan: node ~/Github/axiom-public/agent-skills/skills/social-inbox/social-inbox.mjs\nReview the top 3 items and draft replies if valuable."
}
twitter-api.py (already installed)twitter-replied.json dedup log (already exists)