x-schedule-image-post

Files (1)

• SKILL.md

SKILL.md (first lines)

# X (Twitter) Schedule Image Post with AI Label

Schedule a post on X with an attached image and the "AI-generated" content disclosure label, then verify it appears in the scheduled posts list.

## Prerequisites

- ai-chrome-pilot server is running
- User is logged in to X (if not, run the `x-login` skill first)
- Prefer `HEADLESS=0` for real scheduling flows so the visible composer state can be inspected

## Required inputs

- **Post text**: The text to post
- **Image path**: Absolute path to a local image file (PNG, JPEG, WEBP, or GIF)
- **Scheduled date/time**: Year, month, day, hour, and minute

Optional inputs:

- **Profile name**: Chrome profile to use (default: `default`)
- **AI label**: Whether to enable the "AI-generated" content disclosure (default: `true`)

Default resolution rules:

- Use `default` unless the user explicitly requested another profile
- AI label is enabled by default; disable only if the user explicitly opts out
- If all required inputs are provided, execute directly without an extra confirmation step
- Ask follow-up questions only when required inputs are missing

## Steps

### 1. Verify server is running

```bash
curl -s http://127.0.0.1:3333/health
```

If not running, start it with the resolved profile:

```bash
HEADLESS=0 PROFILE_NAME=<profile_name> npx tsx src/index.ts &
sleep 3 && curl -s http://127.0.0.1:3333/health
```

### 2. Navigate to X home

```bash
curl -s -X POST http://127.0.0.1:3333/goto \
  -H 'Content-Type: application/json' \
  -d '{"url":"https://x.com"}'
```

After `sleep 2`, take a screenshot to confirm the user is logged in.

### 3. Enter post text safely

Use `/snapshot` to find the post text box ref.

```bash
curl -s http://127.0.0.1:3333/snapshot
```

Look for `textbox "ポスト本文"` (Post text), click it, then treat `/act type` only as an initial convenience path.

```bash
curl -s -X POST http://127.0.0.1:3333/act \
  -H 'Content-Type: application/json' \
  -d '{"ref":"<textbox_ref>","action":"click"}'
```

For scheduling that will actually be submitted, prefer the same text-entry strategy as `x-compose-post`:

- connect to the existing Chrome instance with `Playwright over CDP`
- enter text with `page.keyboard.type()`
- verify `[data-testid="tweetTextarea_0"]` exactly matches the intended text before attaching media or opening the schedule dialog

Do not rely on the visible composer alone. X can show an enabled action state while the actual composer DOM is still out of sync.

### 4. Attach the image

Use the `/file-input` endpoint to set the image file on the hidden file input element.

```bash
curl -s -X POST http://127.0.0.1:3333/file-input \
  -H 'Content-Type: application/json' \
  -d '{"files":["<image_path>"]}'
```

After `sleep 2`, take a screenshot to verify the image preview appears in the composer.

### 5. Enable the AI-generated content label

After attaching the image, the toolbar refs will change. Take a fresh snapshot.

```bash
curl -s http://127.0.0.1:3333/snapshot
```

1. Click `button "コンテンツ開示"` (Content disclosure) in the toolbar.

```bash
curl -s -X POST http://127.0.0.1:3333/act \
  -H 'Content-Type: application/json' \
  -d '{"ref":"<content_disclosure_ref>","action":"click"}'
```

2. After `sleep 1`, take a snapshot. Find `switch "AIで生成 ..."` and click it to enable.

```bash
curl -s -X POST http://127.0.0.1:3333/act \
  -H 'Content-Type: application/json' \
  -d '{"ref":"<ai_switch_ref>","action":"click"}'
```

3. Click `button "完了"` (Done) to close the disclosure panel.

```bash
curl -s -X POST http://127.0.0.1:3333/act \
  -H 'Content-Type: application/json' \
  -d '{"ref":"<done_button_ref>","action":"click"}'

From shinshin86/ai-chrome-pilot, MIT.

More from shinshin86/ai-chrome-pilot