Skip to content

mehbylabs/creator-skills / skills/creator-explainer

creator-explainer

Flagged

Turn a topic, article, notes, or document into a narrated explainer video — script, voiceover, per-scene visuals, music bed, captions, final MP4 — using whatever providers the user has, and degrading gracefully (video clips → animated stills → captioned slideshow). Use when the user asks for an explainer, faceless video, narrated video, YouTube video from an article, educational video, listicle video, or "turn this into a video". Not for single clips (creator-video) or product ads (creator-ad).

FileKindSize
SKILL.mdskill md4.1 KB
references/arabic-voice.mdmarkdown7.7 KB
references/audio.mdmarkdown3.4 KB
references/browser-aistudio-browser.mdmarkdown2.8 KB
references/browser-flow-browser.mdmarkdown4.5 KB
references/browser-gemini-browser.mdmarkdown4.8 KB
references/config.mdmarkdown2.5 KB
references/connect.mdmarkdown3.9 KB
references/explainer-structure.mdmarkdown1.4 KB
references/image-prompting.mdmarkdown1.7 KB
references/media-ops.mdmarkdown2.4 KB
references/providers.mdmarkdown7.0 KB
references/video-prompting.mdmarkdown1.9 KB
scripts/connect.pyexecutablescript16.6 KB
scripts/creator.pyexecutablescript71.1 KB
scripts/media.pyexecutablescript15.7 KB

references/browser-gemini-browser.md

View on GitHub
# Playbook: Gemini web app (computer use)

Use only when `creator.py` returned `needs_browser` with `gemini-browser` next, or the user
pinned it. Requires a browser tool (Playwright/Chrome DevTools MCP, Claude in Chrome,
Antigravity's browser agent, or OS-level computer use) and a signed-in Google account.

## Rules

- Navigate by what you see (accessibility snapshot / screenshot), never by fixed coordinates.
  Labels below are hints; the UI changes.
- Never type passwords or solve CAPTCHAs. If signed out, stop and ask the user to sign in.
- One image per chat turn. Start a new chat for unrelated images so earlier context does
  not leak into the result.
- Verify the downloaded file on disk before reporting success.

## Steps

1. Open `https://gemini.google.com/app`. Confirm the account avatar is visible.
2. Start a new chat. If the composer offers a tool picker, choose the image tool
   (often labelled "Create image" or shown with 🍌). Pick the best image model offered
   (Pro / "Thinking" variants give higher quality when available).
3. For reference images, use the composer's upload (+) control and attach the files first.
4. Send the prompt with the ratio stated in digits *and* words, e.g.
   `... Aspect ratio: 16:9 widescreen landscape. No text, no watermark.`
5. Wait for the image to render (typically 10–60 s). Do not send another message meanwhile.
6. Open the image, use its download control (full size if offered). Note the file name.
7. Find the newest file in `browser.downloads_dir` (default `~/Downloads`), then move it:
   ```bash
   python3 scripts/creator.py adopt --src ~/Downloads/<file> --out <requested path>
   ```
   If `adopt` is unavailable, `mv` it and check the real format with `file`.
8. Look at the image. If it misses the brief, refine in the same chat with one change at a time.

### Getting the bytes when driving via Chrome DevTools Protocol

Clicking the download control needs a real mouse event, which raw CDP makes awkward.
Two in-page recipes that work instead (the image `<img>` has `alt=", AI generated"`):

- Canvas (most reliable): draw the img to a canvas and read `toDataURL('image/png')`,
  then base64-decode client-side. The blob URL is same-origin, so the canvas isn't tainted.
- `fetch(img.src)` on the blob URL can return `Failed to fetch` (blob revoked after
  render) — don't rely on it.

The composer is `div[contenteditable="true"][role="textbox"]` ("Enter a prompt for
Gemini"); focus it and `document.execCommand('insertText', …)`; the send button is
`button[aria-label="Send message"]`. No "New chat" button is visible on the app
landing page — sending from it already starts a fresh chat.

### Gotchas seen in the late-2026 UI (verified)

- **Browser zoom ≠ 100% breaks coordinates.** If clicks land ~100 px off what the DOM
  reports, reset zoom (Ctrl+0) and confirm the screenshot height equals
  `window.innerHeight`.
- **An invisible consent overlay swallows every click and key** after the first upload
  ("Creating content from images and files" → Cancel / Agree). If clicks do nothing,
  check `document.elementFromPoint` for `.cdk-overlay-backdrop` and press **Agree**.
- **Uploading files:** the `+` button is "Upload & tools" (or "Upload and tools" on some
  accounts) → menu item "Upload files". Feed local paths with CDP
  `Page.setInterceptFileChooserDialog` + `Page.fileChooserOpened` →
  `DOM.setFileInputFiles({backendNodeId, files})`. DataTransfer injection into
  `input[type=file]` is unreliable.
- **Sending:** Enter in the composer does not always send. The trusted click on the send
  arrow can lag several seconds; a JS `.click()` on `button[aria-label="Send message"]`
  works. Verify the composer emptied before waiting for a result.
- **Generated images may open in a full-screen editor** (Sketch / Text / Save) instead of
  appearing inline. Use its top-bar download icon (direct JPEG, no menu), then press back.
- **Don't navigate with `location.href`** between generations; it detaches the CDP
  execution context. Click the sidebar "New chat" instead.
- The `+` menu also offers **Create music** (Lyria): Length (Short / Standard ≈ 3 min),
  Vocals (Instrumental), Genre; download via "Download track" → "Audio only MP3 track".
- Gemini can **judge audio**: attach several voice takes and ask a native-dialect rubric
  ("rate 1–10 how authentically Tunisian, which words drift"). That's useful for picking a
  TTS voice when the user can't listen.

## Known limits

- Images may carry a visible Gemini watermark depending on the user's plan; SynthID is always present.
- Aspect ratio is prompt-driven; re-ask once if it comes back square.
- "No text" in the prompt is often ignored — the model loves putting signage in scene
  illustrations. This is another reason the skills render real headlines locally.
- Video in the Gemini app is plan-limited and short; prefer `flow-browser` for video.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall