Skills Hub · free skill for bots
Live Layout Watch
Catch layout regressions on a long-running live video: locked regions, edge collisions, and frozen-frame dead air — without stream furniture jargon.
# Live Layout Watch
Built by Dead Brands (deadbrands.co). Outside helper on the original draft: SteTy.
Watch a long-running live video like a producer: protect the locked layout, catch edge collisions, and flag dead air before viewers leave.
## Purpose
Help an outside producer or bot spot **layout regressions** on a continuous live video by comparing fixed on-screen regions over time. This is not Stream-Watch Eyes (that skill is a short checklist of real stream problems to verify). This skill is the ongoing region-lock + collision + freeze sampler you run on a cadence.
## When to use
- You run or monitor a continuous live show with fixed on-screen regions.
- New overlays are being added and may collide with existing elements.
- Viewers report "frozen" or black gaps between loops.
## Limits
- This skill does not restart production software, kill processes, or change production machines.
- Do not publish private credentials, panel passwords, or internal hostnames.
- Frame sampling needs a tool you already have; otherwise mark capture tooling "Planned, not yet available:".
- Aesthetic taste calls stay with the human producer.
- Stay out of KEEP_DRAFT: no house stream furniture vocabulary, no private ingest details.
## Steps
1. Snapshot the locked layout regions (main picture, side panels, bottom banner, chat panel, status strip).
2. Before airing a new element, check it against the lock — enlarging the picture by deleting furniture is a regression.
3. Inspect seams: text under heroes, badges stacking on labels, cards covering the play area.
4. Sample frames on a cadence; flag stretches where pixels stop changing while the show should be live.
5. File a short incident note: time, region, screenshot/hash, suggested owner.
6. Confirm the fix on a fresh sample before closing.
## Examples
Input: Bottom banner title overlaps the score badge after a font change.
Output:
- Collision at bottom-banner / score-badge seam; screenshot hash recorded.
- Revert font size; re-sample 2 minutes clean → closed.