---
name: x-openclaw
description: Use this skill for all X or Twitter tasks in OpenClaw-style environments. Covers setup, authentication, safe command execution, reading timelines, tweets, users, search results, bookmarks, and write actions like post, reply, quote, like, retweet, follow, and bookmark. Prefer this skill when the user wants a single self-contained SKILL.md with no extra references or scripts.
---

# X OpenClaw Skill

Single-file X or Twitter skill for environments that only accept one `SKILL.md`.

## When To Use

Use this skill whenever the user wants to:

- read X timelines, bookmarks, tweets, replies, lists, or user profiles
- search tweets, users, or topics
- post, reply, quote, delete, like, retweet, bookmark, follow, or unfollow
- operate through a terminal agent with minimal extra files

## Tool Assumption

This skill assumes the machine can run the `twitter` CLI from `twitter-cli`.

If `twitter` is missing, install it first:

```bash
uv tool install twitter-cli
```

Upgrade when behavior looks stale or APIs changed:

```bash
uv tool upgrade twitter-cli
```

## Core Rule For Agents

Before any real X action, verify auth first. Do not assume cookies are already available.

```bash
twitter status --yaml >/dev/null && echo "AUTH_OK" || echo "AUTH_NEEDED"
```

If auth is missing, guide the user through one of these:

```bash
twitter whoami
TWITTER_BROWSER=chrome twitter whoami
TWITTER_BROWSER=arc twitter whoami
```

If browser extraction is unavailable, use environment variables:

```bash
export TWITTER_AUTH_TOKEN="<auth_token>"
export TWITTER_CT0="<ct0>"
twitter whoami
```

Then verify:

```bash
twitter status
twitter whoami
```

## Authentication Guidance

Preferred order:

1. Browser cookie extraction
2. Explicit browser selection
3. Environment variables

Common recovery:

- no browser cookies found: ask the user to log into `x.com` in a supported browser, then retry
- read works but write returns 226: full cookies are missing, prefer browser cookie extraction
- `401` or `403`: cookies expired, re-login and retry
- password changed: re-extract cookies

Never ask the user to paste full cookie strings unless there is no safer option.

## Output Preference

For agent workflows, prefer structured output:

```bash
twitter feed --json
twitter tweet "<tweet_id_or_url>" --json
twitter search "AI agent" --json
```

Useful rule:

- use `--json` when piping into `jq`
- use `--yaml` for readable structured output
- use rich output only when the user wants human-readable terminal formatting
- use `-c` for compact LLM-friendly output when token budget matters

## Safe Execution Rules

- avoid bursty write actions
- do not mass-like, mass-follow, or mass-reply without explicit user intent
- prefer read-only commands first, then write actions after confirmation when the action is public or destructive
- treat cookies and session state as secrets
- re-check results after important write operations

## Command Cheatsheet

### Read

```bash
twitter status
twitter whoami
twitter user "<handle>"
twitter feed
twitter feed -t following
twitter bookmarks
twitter search "关键词"
twitter tweet "<tweet_id_or_url>"
twitter show 1
twitter list "<list_id>"
twitter user-posts "<handle>"
twitter likes "<handle>"
twitter followers "<handle>"
twitter following "<handle>"
```

### Write

```bash
twitter post "正文"
twitter post "正文" --image a.jpg
twitter reply "<tweet_id>" "回复内容"
twitter quote "<tweet_id>" "引用内容"
twitter delete "<tweet_id>"
twitter like "<tweet_id>"
twitter unlike "<tweet_id>"
twitter retweet "<tweet_id>"
twitter unretweet "<tweet_id>"
twitter bookmark "<tweet_id>"
twitter unbookmark "<tweet_id>"
twitter follow "<handle>"
twitter unfollow "<handle>"
```

## Recommended Workflows

### Search then inspect

```bash
twitter search "独立开发" --json
twitter show 1 --json
```

### Inspect a tweet thread and replies

```bash
twitter tweet "<tweet_id_or_url>" --json
```

### Inspect a user before following

```bash
twitter user "<handle>" --json
twitter user-posts "<handle>" --json
twitter follow "<handle>"
```

### Post and verify

```bash
twitter post "今天发布一个小更新"
twitter user-posts "<your_handle>" --json
```

## What This Skill Can Do Well

- timeline and bookmark reading
- tweet and reply inspection
- creator and topic discovery
- compact structured output for agent reasoning
- normal posting and engagement workflows
- single-file skill distribution for OpenClaw-like setups

## Known Limits

- write actions are more sensitive than reads
- incomplete cookies may allow reading but break posting
- rate limits and risk controls may interrupt automation
- some endpoints may change with platform updates

## Agent Decision Policy

Follow this order:

1. Check auth
2. Prefer read-only commands to gather context
3. Use structured output when reasoning over results
4. Ask before public or destructive write actions unless the user already requested them clearly
5. Execute writes one by one
6. Re-check results after important actions

## Minimal Examples For OpenClaw

```bash
# auth check
twitter status

# read home timeline
twitter feed --json

# read a tweet
twitter tweet "https://x.com/user/status/1234567890" --json

# search posts
twitter search "AI 工具" --json

# like and reply
twitter like "<tweet_id>"
twitter reply "<tweet_id>" "感谢分享"
```

This skill is intentionally self-contained. If the runtime supports only a single skill document, this file should be enough for normal X or Twitter agent operations.