OffchainLabs/arbitrum-docs / .claude/skills/arbitrum-brand-video-explainers
arbitrum-brand-video-explainers
VerifiedUse when asked to rebrand, restyle, re-pace, resize, or re-render an existing explainer video, when a video's labels are baked-in raster pixels that cannot be edited, when a reviewer says an animation shows the wrong mechanism, or when building a new animated explainer or short looping animated diagram for the Arbitrum docs from a prompt or a doc section.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 18.3 KB |
| examples/nonce-retry-buffer/build.py | script | 16.4 KB |
| examples/pga-rounds/build.py | script | 20.9 KB |
| references/reading-a-source-video.md | markdown | 4.0 KB |
| tools/render.shexecutable | script | 2.9 KB |
references/reading-a-source-video.md
View on GitHub# Reading a source video
How to recover an animation's content from a rendered file, cheaply and without
inventing anything. Every technique here was used to reverse-engineer the worked
example in `examples/pga-rounds/`.
## 1. Probe before you look
```bash
ffprobe -v error -show_entries format=duration,size,bit_rate \
-show_entries stream=codec_type,codec_name,width,height,r_frame_rate,nb_frames,pix_fmt \
-of default=noprint_wrappers=1 in.mp4
```
Note the frame count — every later command addresses frames by index `n`. Check
whether an audio stream exists at all; if the source has none, your output should
not invent one.
## 2. Read many frames in one image
Reading frames one at a time burns a tool call and an image each. The `tile`
filter puts many frames in a single image, so one read covers a dozen states.
```bash
# four frames spread across the run, as a 2x2 grid
ffmpeg -v error -i in.mp4 \
-vf "select='eq(n\,0)+eq(n\,60)+eq(n\,130)+eq(n\,257)',tile=2x2" \
-frames:v 1 -y grid.png
```
Scale down when the grid gets large (`,scale=iw*0.5:ih*0.5`), but stop as soon as
labels stop being legible — an unreadable grid is worse than two readable ones.
## 3. Crop to the state line to recover a state machine
Most explainers carry their state in a fixed header region. Crop to just that
band and tile a long strip: you get the full state sequence in one or two reads
instead of dozens.
```bash
# crop=W:H:X:Y — here a 44px-tall header band, every 10th frame, stacked
ffmpeg -v error -i in.mp4 \
-vf "select='not(mod(n,10))*lt(n,130)',crop=720:44:0:30,tile=1x13,scale=iw*1.5:ih*1.5" \
-frames:v 1 -y status_a.png
```
Scaling **up** matters here — small type that is marginal at 1:1 becomes reliable
at 1.5×. This one technique recovered the complete phase/counter sequence of the
worked example (18 distinct states) from two images.
## 4. Mine the accumulator frames
Animations that build up a history — a ledger, a list of completed items, a
scoreboard — hold most of their data in the **final** frame. Read that first: it
often hands you the whole dataset at once, and earlier frames then only need to
confirm ordering and timing.
In the worked example the "sealed blocks" strip at the end carried four of the
six blocks' full contents; two mid-run frames covered the rest.
## 5. Do not assume one number derives from another
Check before you infer. In the worked example the chips looked like gas units and
the header showed `block gas: N / 150`, so "gas = sum of chips" was the obvious
reading. It is wrong:
| Block | Chips | Sum | Displayed gas |
| ----- | ---------- | --- | ------------- |
| 1 | 31, 9 | 40 | **34** |
| 6 | 21, 12, 11 | 44 | **152** |
The chips are tips; gas is independent and not displayed per transaction. Any
model "fitted" to reconcile them produces numbers that are wrong everywhere.
**The rule:** if a displayed value is not derivable from what is on screen,
transcribe the observed values at the frames you sampled and script those. Record
the frames you sampled it at. Do not fit a curve.
## 6. Infer categorical rules, but state them as inferences
Colour and shape usually encode a category. You can infer the rule, but write
down the evidence and the uncertainty.
Worked example: red chips were tips ≥ 15 (observed reds: 31, 24, 29, 27, 17, 36,
29, 28, 21; observed non-reds: 12, 11, 9, 10, 5, 4, 1). The threshold is bounded
by the data to somewhere in 13–17 — pick one, and say so.
## 7. Write SOURCE.md before writing the renderer
One line per datum, each with the frame number it came from, and an explicit
`UNKNOWN` for anything unreadable. This is the artifact the transcription
contract in SKILL.md refers to. Writing the renderer first turns every gap in
your reading into an invented value, because the renderer needs _something_ in
that slot and you are the one holding the keyboard.
## 8. Verify against your own output
Re-run steps 2 and 3 against the MP4 you produced and diff the result against
`SOURCE.md`. Reading your own SVGs proves nothing — they are the thing under
test.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file). | Stars: 198k | Updated | Claude CodeCodex | Unverified | Install |
| doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |