Stats
Actions
Tags
From Docs from videos
Refresh existing documentation screenshots from a new recording. Use this skill when the user has re-recorded a demo and wants to replace stale screenshots in the docs with fresh ones — e.g. "update the screenshots on the profile page from this new video", "these captures are outdated, here's a new recording", "refresh the variable screenshots from this mp4". It reads each target screenshot's caption from the docs, asks Gemini where that screen appears in the new video, grabs full-resolution candidate frames, and — after you pick — overwrites the images in place (no MDX changes). For documenting brand-new features instead, use video-to-docs; for a raw feature inventory, use extract-product-doc-from-videos.
How this skill is triggered — by the user, by Claude, or both
Slash command
/doc:update-screenshotsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
When a UI is re-recorded, the screenshots in the docs go stale — but the docs
When a UI is re-recorded, the screenshots in the docs go stale — but the docs already say what each image should show (its caption / alt text). This skill uses those captions as the spec: it finds each known screenshot inside the new video, re-grabs it at full resolution, and overwrites the file in place. Because the filename doesn't change, no MDX is touched — the pages just show the fresh images.
docs.json and the
images/ directory).uv + Gemini access). For full-quality
framing on Vertex, set $GEMINI_VIDEO_BUCKET (or pass --gcs-bucket).Two ways, depending on what the user gives you:
images/account-pat.webp).
For each, grep the repo for the filename and read the alt text of the place
it's embedded — that caption is the description of what the shot must show.administration/account/profile).
Read the page, list every image it references with its alt text, and ask the
user (AskUserQuestion) which ones to refresh.Either way you end up with a list of {path, description} targets, where the
description is the existing caption. Don't invent new descriptions — the whole
point is to match what the docs already promise.
Write the targets to a JSON file and run the script with the new video:
cat > /tmp/shot-targets.json <<'JSON'
[{"path": "images/account-pat.webp", "description": "the personal access token field in the Profile section"}]
JSON
uv run "${CLAUDE_PLUGIN_ROOT}/skills/update-screenshots/scripts/update_screenshots.py" <new-video> \
--targets-file /tmp/shot-targets.json --candidates 2
It uploads the video full-quality (GCS on Vertex, else inline), asks Gemini for up
to 2 timestamps per target, grabs those frames full-res into a staging dir, and
prints a manifest.json mapping each target path to its candidate files.
For each target, look at the candidate frames (read the image files). A timestamp can land mid-transition, which is why there are two — choose the clean, fully-rendered one. Recommend the better candidate, but let the user make the call when it matters. If a target got no candidates, Gemini didn't find that screen in the new video — tell the user, don't substitute something wrong.
For each confirmed choice: back up the current image (copy to <name>.bak), then
copy the chosen candidate over the real path in images/. Same filename → the MDX
and docs.json stay untouched.
Restart the dev server (new image bytes are synced at startup, not by hot reload)
and point the user at the affected pages. Once they're happy, remove the .bak
files and the staging dir. Commit only when the user asks.
.bak until the user confirms, and don't
replace a shot Gemini couldn't confidently locate.Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
Applies a firm's KYC/AML rules grid to parsed onboarding records: assigns risk rating, checks required documents, outputs rule outcomes with citations, and routes for escalation.
Generates daily or weekly digests of activity from connected sources (chat, email, docs, tasks, CRM), highlighting action items, decisions, mentions, and project updates.
npx claudepluginhub kapptivate/ktm-docs-jamdesk --plugin doc