Menu
Use when the user asks to create, generate, or render images through the internal media CLI.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Recipe for rendering a code review as a Canvas — compose findings with evidence and actions into a scannable page instead of a long text answer.
Use when the user asks to create a short drama from a prompt — e.g. "生成短剧", "制作微短剧", "make a short drama", "turn this script into a short drama", "ショートドラマを生成", "숏드라마를 만들어". Orchestrates script, storyboard, optional character images, per-shot video clips, and ffmpeg composition through the existing media tools.
Configure Gmail Browser with /connect and use Gmail ConnectorAct actions through the global Puffer browser profile.
Authenticate Lark with /connect (delegated to the official lark-cli), then send/react/reply through the connector and consume incoming Lark messages as a stream. Two connectors mirror Telegram — lark-login (your user account; monitor + act as you) and lark-bot (the app bot; auto-reply). Use lark-cli directly from Bash for chat/user lookup and any API not exposed as a connector action.
Configure SlackApp or SlackLogin with /connect, then resolve Slack users/channels and search/read Slack messages through the internal CLI.
Configure Telegram with /connect, then resolve Telegram user/group ids and search Telegram messages through the internal Telegram CLI.
Based on SOC occupation classification
| name | image-generation |
| description | Use when the user asks to create, generate, or render images through the internal media CLI. |
| allowed-tools | ["Bash"] |
| user-invocable | true |
| disable-model-invocation | false |
| requires-action | true |
Use foreground Bash only. allowed-tools is guidance, not the enforcement boundary; media generation is enforced by the internal tool permission path.
Progress-only or promise-only replies are not completion: after activation, either run imagegen or report the concrete blocker plainly.
imagegen --prompt ... --count ... command for one logical request.--count instead of issuing repeated --count 1 commands.--prompt as literal text unless it names a workspace-relative file; prompt file paths should be passed through --prompt.--prompt-reference only when the request supplies additional prompt context.--aspect, --purpose, or --retry-from-error-json only when the request or retry context requires them.--provider <id> --model <id> together — both are required, or pass neither. Omit them to use the configured default provider/model.