Gog
一个面向 Dev Tools 场景的 Agent 技能。原始说明:Google Workspace CLI for Gmail, Calendar, Drive, Contacts, Sheets, and Docs.
name: x-search-oauth
description: "Search X/Twitter with OAuth-backed xAI x_search; includes optional xso CLI companion guidance."
homepage: https://github.com/LeoStehlik/x-search-oauth
metadata:
openclaw:
emoji: "X"
requires:
plugins: ["xai"]
tools: ["x_search"]
install:
kind: "node"
package: "x-search-oauth"
bins: ["xso", "x-search-oauth"]
label: "Install xso CLI companion (npm)"
version: "0.2.3"
Use when the user asks to search X/Twitter, inspect posts, find X trends, monitor AI/tech chatter, look up a post/thread, gather X citations, or use the x-search-oauth / xso CLI.
Inside OpenClaw, use the native x_search tool exposed by the bundled xai plugin when available.
Do not ask for XAI_API_KEY.
Do not use unofficial API-key-only X search skills when native x_search is available.
Do not use scraping paths.
If x_search is unavailable inside OpenClaw, tell the user that the bundled xAI plugin must be enabled and signed in. Keep setup instructions minimal unless the user asks for exact commands.
The same GitHub repo also ships xso, a standalone Node.js CLI for human terminal use. ClawHub installs this skill; OpenClaw can also offer the xso CLI as an optional Node companion binary via the skill metadata.
For terminal use:
xso auth
xso "AI coding agents" --from-date YYYY-MM-DD
The CLI performs xAI device-code OAuth directly and does not require OpenClaw Gateway. OAuth state is stored under the user's config directory, normally ~/.config/x-search-oauth/auth.json.
Prefer several narrow searches over one vague broad search.
Examples:
AI agents coding agents OpenClaw Claude Code Codex
Grok OpenClaw xAI agents
local AI Ollama llama.cpp vLLM Apple Silicon
AI coding agents production deployment evals
browser agents memory RAG autonomous agents
For account-specific checks, use handle filters:
allowed_x_handles: ["openclaw"]
query: "OpenClaw 2026.5.19 xAI login headless"
For freshness, use date filters:
from_date: "YYYY-MM-DD"
to_date: "YYYY-MM-DD"
Use image/video understanding only when the user needs media interpretation; otherwise keep it off for speed.
x_search.For trend reports:
**Top Signal**
- [Observed X activity]. Why it matters: [practical implication]. [citation]
**Watchlist**
- [Weaker but relevant item]. [citation]
**Noise / Ignore**
- [Loud but unsupported/stale item].
**Tool Note**
- x_search was [fast/slow/thin/rich]; note failures or retries.
For exact post/thread lookups, lead with the direct answer, then cite the post URLs and note any uncertainty.