yikart/AiToEarn / skills/doubao-skill-aitoearn/skills/byted-mediakit-media-understand
byted-mediakit-media-understand
VerifiedUnderstand videos, audio and images with MediaKit MCP tools. Use for media summaries, visual event searches, dialogue analysis, sound descriptions, and timestamped evidence for editing. Use sampled-image understanding, ASR and audio understanding; submit video to an understanding model only when the user explicitly requests direct video understanding. Refine only the intervals needed to answer the question.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.6 KB |
| references/mcp.md | markdown | 3.2 KB |
| references/sampling.md | markdown | 3.1 KB |
references/mcp.md
View on GitHub# MCP reference
## Tool mapping
Use the MediaKit tools exposed by this environment. Do not install or run `mediakit-cli`.
| Capability | Tool and arguments |
| --- | --- |
| Frame extraction | `extract_frames` with `video_url` and sampling fields |
| Image / audio / video understanding | `media_understanding` with `media_url` and `prompt` |
| Speech evidence / subtitles | `asr_subtitles` |
| Full transcript | Follow `../../byted-mediakit-asr/SKILL.md` |
| Task status | The environment resolves submitted task IDs automatically |
If a capability is unavailable, report the gap. Do not substitute a prohibited high-level analysis workflow.
## Understanding
Submit one image, audio, or video asset per call. Submit video only when the user explicitly requests direct video understanding:
```json
{
"media_url": "<reference accepted by the MCP schema>",
"prompt": "<question and source evidence context>"
}
```
For image and audio inputs, omit `manual_option` and `scene`. For a contact sheet, provide the verified tile order, source timestamps, coverage gaps and question. Ask for visible findings tied to frames, with interpretations distinguished from observations. For audio, specify the relevant music, sound effects or ambience, source interval, and approximate time ranges; do not ask it to infer visual causes. Use `asr_subtitles` separately for exact speech.
Direct video input requires an explicit user request for direct video understanding; do not add it alongside extraction and ASR for speed, story context or highlight search. Tool support for video and `scene=editing` do not authorize this path or guarantee precise timestamps. For explicitly requested video sound analysis, use `manual_option.need_audio=true` when supported; omit `scene` for sound analysis.
## Extraction
Use typed JSON values. Example arguments:
```json
{
"video_url": "<service-accessible video reference>",
"snapshot_type": "TimeInterval",
"time_interval": 3,
"enable_sprite": true,
"sprite_rows": 4,
"sprite_cols": 4,
"snapshot_limit": 3,
"scale_long": 384
}
```
Choose values for the actual source; this example permits up to 48 tiles and does not guarantee complete coverage. Apply the shared sampling reference, including readable tiles and reliable time mapping. For refinement, use `snapshot_type=SpecifiedTime`, a numeric `specified_time` array and `enable_sprite=false` when supported. Send returned images to image understanding rather than resubmitting the original video.
## Inputs and task lifecycle
Use media references accepted by the MCP schema and runtime resolver. Do not send Base64, data URLs, local paths, or platform IDs as model parameters. If no suitable accessible reference is available, report the missing input capability.
Retain the actual task ID from submission and query the same task until terminal. If submission has no usable task ID, inspect the actual response and report the submission problem without inventing an ID or blindly retrying. A timeout or empty content is not a reason to submit the operation again. Keep source/modality associations, retain usable partial evidence, and report missing analysis. Only summarize content actually returned or inspected.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |