media-analysis

v2026.09.24

Analyze videos, images, or carousels with visible and audible evidence. Answer visual questions, inspect frames, or compare references shot by shot when requested.

GitHub
安装命令
npx skhub add postplusai/media-analysis
Markdown
SKILL.md

Media Analysis

Analyze the user's link, local file, or visible images and deliver readable findings with the actual visual evidence. Choose the necessary work internally.

For multi-reference comparison or requested detailed shot-by-shot evidence, read multi-source shot analysis. This branch keeps sources independent before synthesis; ordinary analysis skips it.

Choose the work

  • Video understanding: use postplus media analyze video-analysis for actions, sound, dialogue, and development over time. Pass the original source; the CLI owns acquisition and transfer. Read execution and recovery.
  • Readable images: view them directly. No download, upload, or Gemini call is needed for images already accessible in the conversation or locally.
  • Native image posts: use postplus media prepare --source <url>, then inspect the ordered originals. Read image evidence.
  • Screenshots, text, color, or layout: inspect sufficient images or local video frames without purchasing full video analysis. Read video frames.

An MP4 made from stills remains a video with a real timeline. Preparation must return an actual local file before you treat it as downloaded evidence. Keep multiple sources, failures, and outputs separate. Answer follow-ups from existing reports and evidence without another paid analysis.

Video command

<!-- BEGIN GENERATED EXECUTION EXAMPLE -->
postplus media analyze video-analysis \
  --video ./reference.mp4 \
  --output ./result.md

Follow the CLI's structured result and reported next action; do not infer recovery from free-text messages. Wait for explicit user approval when requested; an action does not authorize spending, publishing, or overwriting. Resume the same operation through its returned checkpoint or action; never resubmit uncertain work, repeat exhausted recovery, or switch providers to bypass failure.

<!-- END GENERATED EXECUTION EXAMPLE -->

Omit --prompt for the default shot table and production notes described in video report. Supply it for the user's specific question or requested format; do not append an unrelated full report. video-analysis is the CLI capability key, not another Skill. Use postplus media schema --json only when constructing or repairing an unknown hosted media request shape; keep the generated command above for the normal analysis path.

Deliver

Embed inspected images or real screenshots in the conversation or Markdown, with source, original page index or measured video timestamp, and the observation each supports. Paths or folder links alone are not visual delivery. Never replace missing evidence with a cover, generated image, or guessed description.

For a long report, show representative frames in chat and link the full Markdown. Preserve returned model Markdown unchanged; add Agent-viewed screenshots and any corrections in a separate evidence note. Save attachments beside the report using relative links, verify those links, and retain them after temporary cleanup. Never delete user files or delivered evidence.

State missing pages, unreadable text, unavailable audio, and inspected subsets. A few frames do not prove full audio/video coverage. Distinguish visible facts, captured wording, interpretation, and claims that need external verification.

Recover

Report observed progress while waiting. Let the CLI perform bounded recovery; after a disconnected wait use its emitted poll or checkpoint command for the same task, account, and environment. Do not restart an uncertain paid analysis. A content-quality concern or rewording request does not authorize a paid redo.

For an unambiguous local parameter rejection before remote work starts, use the command's --help and existing user input for one correction. If submission may have started, follow only its emitted recovery command. Use postplus doctor --skill media-analysis for readiness diagnostics after a readiness failure. Preserve task identity if submission may have happened. Honor quote challenges using existing explicit cost authorization when sufficient; otherwise ask for approval, then use the provided confirmation/retry commands.

For a terminal failure, give its stage, reason, operation id when present, cost status, and concrete next step. Unknown results do not mean no charge. Deliver useful partial evidence with its limits; never claim unseen material was analyzed.

发现
标签

此技能尚未发布标签。

版本
最新版本元数据

版本

v2026.09.24

发布时间

2026年9月24日

分类

未分类

许可证

NOASSERTION

源路径

skills/40-creative/media-analysis

默认分支

main

最新提交

51a1c76

Tree SHA

ee441c2