video-transcripts - SKILL.md Agent Skill

name: video-transcripts description: Generate structured video transcripts from local files or video URLs using Gemini Files API. Use when a GitHub or Linear tracker item, comment, or attachment includes a screen recording, .mov, .mp4, or tracker-hosted video and you need a block instead of hand-written notes. disable-model-invocation: true

Video Transcripts

Quick Start

Run the helper once per relevant video:

bash .agents/skills/video-transcripts/scripts/generate_video_transcript.sh \
  "https://uploads.linear.app/.../screen-recording.mov" \
  --title "PDF preview hyperlinks trigger leave-page modal"

Or for a GitHub attachment:

bash .agents/skills/video-transcripts/scripts/generate_video_transcript.sh \
  "https://github.com/user-attachments/assets/..." \
  --title "Slash menu loses selection after confirm"

Or for a local file:

bash .agents/skills/video-transcripts/scripts/generate_video_transcript.sh \
  "/absolute/path/to/video.mov" \
  --title "Preview hyperlink exits workflow"

For auth-gated Linear uploads, the helper automatically retries with cookies from the local Linear desktop app on macOS. For private GitHub asset URLs, it retries with GITHUB_TOKEN, GH_TOKEN, or gh auth token when available.

Use This When

A GitHub or Linear issue, PR, or comment includes a screen recording.
An attachment URL points to uploads.linear.app.
An attachment URL points to a GitHub attachment or private GitHub asset host.
You need timeline-style transcript lines, not a vague summary.
You want the result in this exact XML shape:

<video-transcripts>
<video-transcript title="...">
[00:00] (...)
</video-transcript>
</video-transcripts>

Workflow

Run the helper once for each relevant video.
Give each run a short, bug-focused --title.
For tracked work, the canonical shared cache should live in the tracker next to the evidence it describes.
If the video is in the issue or PR body, use a top-level tracker comment.
If the video is in a Linear comment, post the transcript cache as a reply to that specific comment.
If the video is in a GitHub issue or PR comment, use one dedicated top-level cache comment for that source comment's video set. GitHub has no replies, so keep cache comments separated by source container instead of merging unrelated comment videos together.
Cache comment or reply body should start with a transcript source link, then the timestamp lines:

[[Transcript](link to video or comment if not available)]
[00:00] (...)

When caching helper XML, strip the <video-transcripts> and <video-transcript ...> wrapper lines and paste only timestamp lines after the source link.
If there are multiple videos in the same source container, repeat the [[Transcript](...)] source link plus timestamp lines for each video.
Do not hand-write or paraphrase video behavior when the helper can run. Use the actual transcript output.
Link [[Transcript](...)] to the video URL when available. If the video URL is not available or is unstable, link to the source comment that contains the video.
For signed tracker-hosted URLs like uploads.linear.app, strip the query string so a new signature does not invalidate an otherwise valid cache entry.
Do not add decorative metadata like title to cached transcript entries unless a later workflow truly needs it.
Before re-transcribing for tracked work, match cache entries by source container first:
- one cache comment for issue or PR body videos
- one Linear reply per comment containing video(s)
- one dedicated GitHub cache comment per issue or PR comment containing video(s)
If the matching cache already covers the current normalized video keys for that source container, reuse it. Only transcribe missing or new video evidence.
Do not use docs/ for raw tracker transcript cache. That is durable repo knowledge space, not raw issue evidence.
If you invoke this through codex exec, prefer -o <file> so the final XML is captured without CLI progress chatter.

Output Contract

Return XML, not Markdown.
Use one [MM:SS] (...) line per observed action or system response.
Quote visible UI text when legible.
Describe only visible actions, screen changes, and audible speech if present.
Do not invent hidden state, motives, or implementation details.
Prefer concise, high-signal lines over per-keystroke sludge.

Model Strategy

The helper defaults to gemini-3.1-flash-lite-preview for cost efficiency.

If that output is malformed, too thin, or obviously noisy, it retries with gemini-3-flash-preview.

For Gemini 3 models, the helper forces minimal thinking so output budget goes to the transcript instead of hidden reasoning.

Override with:

VIDEO_TRANSCRIPTS_MODEL=gemini-3-flash-preview bash .agents/skills/video-transcripts/scripts/generate_video_transcript.sh ...

Or:

bash .agents/skills/video-transcripts/scripts/generate_video_transcript.sh ... \
  --model gemini-2.5-flash

Notes

The helper accepts a local file path or remote URL.
For uploads.linear.app URLs, it first tries LINEAR_COOKIE_HEADER, then LINEAR_COOKIES_DB, then falls back to the local Linear desktop cookie store at ~/Library/Application Support/Linear/Cookies.
For GitHub asset URLs, it first tries GITHUB_TOKEN, then GH_TOKEN, then gh auth token, then falls back to an unauthenticated download.
It looks for GEMINI_API_KEY, then GOOGLE_API_KEY.
If neither is set, it tries ~/.bash_profile before failing.
Use --debug-dir <dir> when you want request and response artifacts saved.
Quality gate rejects obvious low-signal transcripts such as repeated partial typing runs and falls back automatically.