← Files Creative ClawARCHIVED FILE
skills/creativeclaw/references/tool-catalog.md
13.3 KB · Oct 5, 2026 · 12:04 UTC
# Tool catalog
Use this as routing guidance. Tool availability varies by client; never recite the catalog to the user or invent a missing tool.
Older clients may still show retired tools. Use `search_examples({ id })` instead of `get_example`, `manage_character({ id, delete: true })` instead of `delete_character`, and `generate_music` or `generate_sound_effect` instead of `generate_audio`.
## Models and generation
| Tool | Use |
| ------------------ | ------------------------------------------------------------------------------------------------------------------------- |
| `list_models` | Discover current image, video, speech, and audio models. Filter by category. |
| `get_model_params` | Read the chosen model's actual schema, defaults, enums, and extra fields. |
| `generate_image` | Generate an image or edit a source supplied as `image_url`. Additional model-specific references normally go in `extras`. |
| `generate_video` | Generate, animate, extend, retake, reframe, edit, or drive video according to model capability and `operation`. |
| `generate_speech` | Generate speech. Supports model-specific voices, delivery controls, reference audio, and Characters. |
| `generate_music` | Generate scores, music beds, stings, jingles, themes, or songs. `creativeclaw-generate-music` picks the model. |
| `generate_sound_effect` | Generate sound effects, ambience, Foley, transitions, impacts, or textures with ElevenLabs. |
Use `list_models` before relying on a remembered model ID. Use `get_model_params` before sending `extras`, reference arrays, resolution, duration, or operation-specific fields. Route speech to `generate_speech`, music to `generate_music`, and sound effects to `generate_sound_effect`.
For a model comparison, call `generate_image` once per selected model with the same prompt and settings, then present the results together.
## Curated examples
| Tool | Use |
| ----------------- | ------------------------------------------------------------------------------------------------------------------- |
| `search_examples` | Browse summaries by query, output type, exact model ID, tags, and cursor. Pass `id` (id or slug) to load one example in full. |
Search only when the user asks for examples, prompt ideas, or the catalog; see [examples guidance](workflows/examples.md). Do not search before every generation. Variations and user-supplied style references stay with the generation skill.
## Deterministic HTML rendering
| Tool | Use |
| ------------------- | -------------------------------------------------------------------------------------------------------------- |
| `render_html_image` | Render a deterministic HTML/CSS layout to a PNG via Chromium. |
| `render_html_video` | Queue a HyperFrames-backed HTML/CSS/JS motion render; resolve the final video URL with `check_job`. |
| `list_templates` | Find saved reusable layouts by name. |
| `create_template` / `update_template` | Save or change a reusable parameterized layout. |
| `render_template` | Render a saved template with provided values. |
The image and video renderers are explicit-only choices. Use them when the user asks for HTML/CSS, HyperFrames, code-driven rendering, supplies HTML, or explicitly accepts the method. `render_html_image` can also make an exact text watermark graphic with a transparent background for `merge_media` `overlay_images`. A generic poster, social card, overlay, intro, or outro request is not sufficient by itself.
## Themes
| Tool | Use |
| -------------- | ----------------------------------------------------------------------------------------------------- |
| `list_themes` | Discover theme names and the default theme. |
| `get_theme` | Fetch the default or named theme, including structured data and reference images. |
| `update_theme` | Open the visual editor with `interactive: true`, or directly create/update data and reference images. |
For conversational setup or edits, prefer `update_theme({ interactive: true })`. For exact programmatic changes, fetch first and send the smallest direct update. Theme `data` shallow-merges; `images` replaces the full reference-image array.
## Assets and imports
| Tool | Use |
| ----------------------------------- | ------------------------------------------------------------------------------------------------------ |
| `search_assets` | Find recent or matching media by type, query, tags, exact name, or source. |
| `update_asset` | Set a unique name, replace tags, or update a description. |
| `delete_asset` | Permanently delete an asset; it cannot be restored. Only when explicitly requested. |
| `load_image` | Display an image URL inline when exposed. |
| `import_chatgpt_media` | Convert one native ChatGPT attachment into a durable Creative Claw asset when exposed. |
| `import_media` | Open the picker when the user still needs to choose a file. |
| `get_upload_url` + `confirm_upload` | Upload local bytes from clients that can perform an HTTP PUT. |
| `upload_asset` | Copy a public, directly downloadable URL into the library. Requires `url`, `content_type`, and `type`. |
Read `platform-upload.md` before choosing an import route.
## Characters, voices, and films
| Tool | Use |
| ----------------------------------------- | ------------------------------------------------------------------------------------------------- |
| `manage_character` | Create or update a Character (description, reference image, stock voice). `delete: true` with `id` permanently deletes it and its saved voices, only on explicit request. |
| `list_characters` | Find reusable Characters, their IDs, and saved voice options. |
| `design_voice` | Create a new synthetic voice from a description, no recording needed. ElevenLabs by default; `provider: "google"` for Google. Returns three auditions. Save the pick with `design_voice({ action: "save", preview_id, character_id \| character_name })`. At the saved-voice limit, pass `replace_voice_option_id` to replace an existing designed voice. |
| `clone_voice` | Clone a consenting speaker's recording with Cartesia (default) or ElevenLabs (`provider: "elevenlabs"`). Attaches to `character_id`, or creates a voice-only Character when none is given. |
| `create_film_project` | Create a multi-shot Film project. |
| `update_film_project` | Save script, shots, storyboards, clips, audio, and approval state. |
| `get_film_project` / `list_film_projects` | Inspect Film projects. |
| `assemble_film` | Concatenate every approved shot clip into a first cut and optionally overlay one project narration track. |
## Media processing
| Tool | Use |
| ----------------------- | ----------------------------------------------------------------------------------------- |
| `remove_background` | Remove an image or video background. |
| `upscale_media` | Upscale image or video. |
| `trim_video` | Cut one time range. |
| `cut_and_reframe_video` | Cut and reorder several chosen ranges of a workspace video in one render, reframe each for any output size, and optionally burn source-timed captions. Does not pick highlights or track faces. |
| `scale_video` | Resize, crop, or pad video. |
| `add_subtitles` | Burn captions into video. |
| `extract_frames` | Extract one or more still frames from video. |
| `merge_media` | Concatenate media, put audio on a video, burn transparent image watermarks onto finished video (`overlay_images`), or compose timed images and clips with optional audio (`compose_video`). See [assembly guidance](media-assembly.md). |
| `transcribe` | Timed transcript from `audio_url` or `video_url`. `video_url` also accepts YouTube, Google Drive, and public video pages (TikTok, Instagram, X/Twitter, Facebook, Vimeo, Twitch, Dailymotion, Loom, Reddit), best effort; import the file if a page cannot be fetched. |
| `isolate_audio` | Remove noise, music, and reverb from a voice recording. Queued; resolve with `check_job`. |
Audio on video:
- `merge_media` `merge_audio_video` replaces the clip's audio by default and ends at the shorter input.
- With `audio_mode: "mix"` it keeps the clip's sound and layers the new track over it at the video's length. `original_volume` and `added_volume` (0–1) set each level; about 0.3 keeps music under speech.
- `compose_video` layers one audio track over the clips at full level, and the result can run past the last clip.
## Account and billing
Use `manage_account` to view the connected account, balance, recent generations, and credit activity, including charges and refunds. Open `overview` for the balance, `activity` for a specific generation's recorded cost, or `settings` for generation preferences. The tool is read-only.
Purchases and plans are handled on the Creative Claw website, not in chat. Do not start, price, or recommend a purchase; if asked, say so and share only an account link a tool returned.
Read [account guidance](workflows/account.md) to match jobs to charges, distinguish gross charges/refunds/net cost, and handle missing or older records. Use `estimate_generation` for future quotes, not proof of what a past generation cost. A balance question alone is not a feedback submission request.
## Jobs and estimates
| Tool | Use |
| --------------------- | -------------------------------------------------------------------- |
| `check_job` | Resolve queued work by `job_id` when a completed result is required. |
| `estimate_generation` | Estimate one planned generation, check the current balance, and return affordable H3 alternatives when needed. The final cost is confirmed after generation finishes. |
| `get_credits_balance` | Check the balance when exposed. |
## Product feedback
| Tool | Use |
| ----------------- | ------------------------------------------------------------------------------------------------------------ |
| `submit_feedback` | Report a bug, missing feature or model, confusing flow, generation-quality issue, or explicit user praise. |
Use `source: "agent"` for friction observed during a task and `source: "user"` when relaying the user's own feedback. Send one concise, specific report only when the user asks or approves; a complaint or account question alone is not authorization. Completed playable video generations are charged even if the user is disappointed. Feedback helps improve future generations and does not itself trigger a refund. For critical issues, users can also contact support@creativeclaw.co. Read [feedback guidance](workflows/feedback.md) before reporting.
## Metadata conventions
- Use concise unique names such as `acme-launch-hero-v2`.
- Use lowercase stable tags such as `acme`, `launch-2026`, `approved`, `reference`, `character-mira`.
- `update_asset.tags` replaces the current tag array; include every tag that should remain.
- Search before generating a duplicate and before assigning a name that must be unique.
SHA-256: 11d9accdf02e32abc8551fe89544bea9cf6cbc6db4ce10ce3df76d682867c299