agentleFS
JuanMarchettoJuanMarchetto/agent-skills/skills/webreel/SKILL.md

webreel

Skill5 starsChanged 8 months agoMITView on GitHub

Create and record scripted browser demo videos with webreel. Generates MP4, GIF, or WebM recordings with cursor animation, keystroke overlays, and sound effects from a JSON config. Use when: record a demo, create a browser video, edit a webreel config, generate a screen recording, preview a demo, product video, marketing video, record website, work with webreel in any way.

  • Installs packages

1 other project uses a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

27 rules. About 0 are common, 7 are distinctive, and 20 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 20

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the project already has webreel in its dependencies, skip this step.

    line 23
  2. Very specific:

    If a recording fails with "No inspectable targets" or similar browser errors, the issue is almost certainly in the webreel config (wrong …

    line 42
  3. Distinctive:

    When the user asks to record a demo, follow these steps:

    line 71
  4. Distinctive:

    If the user provides a URL but no specific steps, suggest a sensible default flow: navigate to the page, wait for load, scroll down slowly, …

    line 80
  5. Distinctive:

    Run steps in a visible browser without recording.

    line 123
  6. Distinctive:

    Check config for errors without running anything.

    line 150
  7. Very specific:

    Use preset names as string values for viewport, or specify { width, height }:

    line 250
  8. Very specific:

    Always set waitFor on a video to ensure the page is ready before steps run.

    line 326
  9. Very specific:

    Use delay on individual steps to control pacing between actions.

    line 327
  10. Very specific:

    Use --watch during development for automatic re-recording on config changes.

    line 328
  11. Very specific:

    Use composite to iterate on theme/overlay settings without re-recording.

    line 329
  12. Very specific:

    Use --verbose to debug step execution.

    line 330
  13. Very specific:

    Use --dry-run to inspect the fully resolved config (includes, env vars, defaults).

    line 331
  14. Very specific:

    Use zoom to scale up small UIs for readability in the recording.

    line 332
  15. Very specific:

    Start with preview to verify steps work before committing to a full recording.

    line 333
  16. Very specific:

    Run npx webreel record --verbose to see which step fails

    line 347
  17. Very specific:

    Check waitFor — if the selector doesn't exist on the page, the browser times out

    line 348
  18. Very specific:

    Run npx webreel install --force to re-download binaries if they're corrupted

    line 349
  19. Very specific:

    If using a custom CHROME_PATH, verify the binary exists and is executable

    line 350
  20. Very specific:

    Add a wait step before interacting: { "action": "wait", "selector": ".hero-section", "timeout": 10000 }

    line 362
  21. Very specific:

    Use npx webreel preview --verbose to watch the browser and see what's on screen

    line 363
  22. Distinctive:

    Verify selectors with browser DevTools on the target URL

    line 364
  23. Very specific:

    If the page has lazy-loaded content, add a scroll step to trigger loading first

    line 365
  24. Very specific:

    Add delay to individual steps: { "action": "click", "text": "Submit", "delay": 500 }

    line 372
  25. Very specific:

    Use clickDwell to add a pause before clicks (makes cursor movement feel natural)

    line 374
  26. Very specific:

    Run npx webreel record --frames to save raw JPEGs — check if frames are captured

    line 382
  27. Distinctive:

    Verify the URL is accessible (not behind auth, VPN, or localhost that isn't running)

    line 383

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

webreel

webreel records scripted browser demos as MP4, GIF, or WebM with cursor animation, keystroke overlays, and sound effects. You define steps in a JSON config, and webreel drives headless Chrome, captures frames, and encodes with ffmpeg.

Installation

Install webreel as a project dependency so the version is pinned in the lockfile. This ensures deterministic recordings across machines and CI.

npm install webreel

If the project already has webreel in its dependencies, skip this step.

Prerequisites

webreel requires Chrome and ffmpeg, but you do NOT need to install them manually. Both are automatically downloaded to ~/.webreel on first run if not already present. Do not install Chrome or Chromium via puppeteer, playwright, or any other tool. webreel manages its own browser.

To download dependencies explicitly, or to fix corrupted/broken binaries:

npx webreel install
npx webreel install --force   # delete cached binaries and re-download

To override the auto-downloaded binaries, set these environment variables:

  • CHROME_PATH - path to a Chrome or Chromium binary (used for preview)
  • CHROME_HEADLESS_PATH - path to a chrome-headless-shell binary (used for recording)
  • FFMPEG_PATH - path to an ffmpeg binary

If a recording fails with "No inspectable targets" or similar browser errors, the issue is almost certainly in the webreel config (wrong waitFor, missing element, timing), not a missing browser. Check the config and use --verbose to debug.

.gitignore

The .webreel directory is created at the project root during recording (frames, intermediate files). Add it to .gitignore:

.webreel

Quick start

# Scaffold a config
npx webreel init --name my-demo --url https://example.com

# Edit webreel.config.json with your steps

# Preview in a visible browser (no recording)
npx webreel preview my-demo

# Record the video
npx webreel record my-demo

npx resolves to the locally installed version when webreel is in devDependencies. Output lands in videos/ by default (configurable via outDir).

Agent Workflow

When the user asks to record a demo, follow these steps:

  1. Check for config — look for webreel.config.json (or .ts/.js variants) in the project root
  2. Initialize if missing — if no config exists, run npx webreel init --name <name> --url <url> to scaffold one
  3. Edit config with steps — add steps to the videos map based on what the user wants to demonstrate (clicks, typing, navigation, etc.)
  4. Preview first — run npx webreel preview <name> to verify steps work in a visible browser before recording
  5. Record — run npx webreel record <name> to produce the final video
  6. Iterate — if the user wants to adjust timing, theme, or overlays, edit the config and use npx webreel composite <name> to re-apply overlays without re-recording

If the user provides a URL but no specific steps, suggest a sensible default flow: navigate to the page, wait for load, scroll down slowly, highlight key UI elements with clicks or hovers.

When to Use webreel vs Other Tools

ToolBest ForKey Difference
webreelScripted demos with polish (cursor animation, keystroke overlays, sound effects)Deterministic, config-driven, produces production-quality video
Playwright recordingQuick screen captures for bug reports or docsFast but no cursor animation, no overlays, no post-processing
demo-pipelineNatural language to automated demoHigher-level abstraction, less control over individual steps
Agent browser (Playwright MCP)Live interaction and testingReal-time browsing, not optimized for video output

Choose webreel when: the output needs to look polished (marketing, onboarding, product demos, social media), steps are known in advance, and you want repeatable recordings.

CLI commands

init

Scaffold a new webreel.config.json.

webreel init
webreel init --name login-flow --url https://myapp.com
webreel init --name hero -o hero.config.json

Flags: --name (video name), --url (starting URL), -o, --output (output file path).

record

Record one or more videos.

webreel record                        # all videos in config
webreel record hero login             # specific videos by name
webreel record -c custom.config.json  # custom config path
webreel record --watch                # re-record on config change
webreel record --verbose              # log each step
webreel record --dry-run              # print resolved config only
webreel record --frames               # save raw JPEGs to .webreel/frames/

preview

Run steps in a visible browser without recording.

webreel preview
webreel preview hero --verbose

composite

Re-apply overlays (cursor, HUD, sfx) to existing raw video without re-recording. Useful for tweaking theme settings.

webreel composite
webreel composite hero

install

Download Chrome and ffmpeg to ~/.webreel. Also use this to fix corrupted or broken binaries.

webreel install
webreel install --force  # delete cached binaries and re-download

validate

Check config for errors without running anything.

webreel validate
webreel validate -c custom.config.json

Config structure

Config files are auto-discovered as webreel.config.json (or .ts, .mts, .js, .mjs). Use -c to specify a custom path.

Top-level fields

FieldDefaultDescription
$schema-"https://webreel.dev/schema/v1.json"
outDir"videos/"Output directory for rendered videos
baseUrl""Base URL prepended to relative video URLs
viewport1080x1080Default viewport { width, height }
theme-Cursor and HUD overlay theme
sfx-Sound effect settings
include-Array of step file paths prepended to all videos
defaultDelay-Default delay (ms) appended after each step
clickDwell-Cursor dwell time (ms) before a click

Per-video fields

Each entry in the videos map supports:

FieldDefaultDescription
urlrequiredURL to open (absolute or relative to baseUrl)
viewportinheritedOverride viewport { width, height }
zoom-CSS zoom factor
waitFor-Selector or text to wait for before starting steps
output"<name>.mp4"Output path (.mp4, .gif, .webm)
thumbnail{ time: 0 }Thumbnail config, or { enabled: false }
includeinheritedStep files to prepend
themeinheritedOverride theme
sfxinheritedOverride sound effects
defaultDelayinheritedOverride default delay
clickDwellinheritedOverride click dwell
fps60Frame rate
quality80Encoding quality (1-100)
stepsrequiredArray of step objects

Videos map

Videos are keyed by name in the config:

{
  "videos": {
    "hero": { "url": "...", "steps": [...] },
    "login": { "url": "...", "steps": [...] }
  }
}

Record specific videos by name: webreel record hero login.

Step types

Each step has an action field. Most steps accept optional label, delay (ms after step), and description fields.

ActionKey fieldsPurpose
pausemsWait for a duration
clicktext or selector, within, modifiersClick an element
typetext, selector, within, charDelayType text into an input
keykey, targetPress a key combo (e.g. "cmd+s")
dragfrom, to (element targets)Drag between two elements
scrollx, y, selectorScroll the page or an element
waitselector or text, timeoutWait for an element to appear
moveTotext or selector, withinMove cursor to an element
navigateurlNavigate to a new URL
hovertext or selector, withinHover over an element
selectselector, valueSelect a dropdown value
screenshotoutputCapture a PNG screenshot

For full field details on every step type, see steps-reference.md.

Element targeting

Many steps target elements using these fields:

  • text - match by visible text content
  • selector - match by CSS selector
  • within - narrow the search to a parent matching this CSS selector

You can use text or selector (not both). within is optional and scopes the search.

{ "action": "click", "text": "Submit" }
{ "action": "click", "selector": "#submit-btn" }
{ "action": "click", "text": "Submit", "within": ".modal" }

Viewport presets

Use preset names as string values for viewport, or specify { width, height }:

desktop (1920x1080), desktop-hd (2560x1440), laptop (1366x768), macbook-air (1440x900), macbook-pro (1512x982), ipad (1024x1366), ipad-pro (834x1194), ipad-mini (768x1024), iphone-15 (393x852), iphone-15-pro-max (430x932), iphone-se (375x667), pixel-8 (412x915), galaxy-s24 (360x780).

Theme

Customize cursor appearance and keystroke HUD:

{
  "theme": {
    "cursor": {
      "image": "./cursor.svg",
      "size": 32,
      "hotspot": "center"
    },
    "hud": {
      "background": "rgba(30, 41, 59, 0.85)",
      "color": "#e2e8f0",
      "fontSize": 48,
      "fontFamily": "\"SF Mono\", monospace",
      "borderRadius": 12,
      "position": "top"
    }
  }
}
  • cursor.image - path to a custom cursor SVG or PNG
  • cursor.size - cursor size in pixels
  • cursor.hotspot - "top-left" (default) or "center"
  • hud.position - "top" or "bottom"

Common patterns

Shared steps via include

Factor out reusable step sequences (e.g. dismissing a cookie banner) into JSON files:

// steps/dismiss-banner.json
{
  "steps": [
    { "action": "wait", "selector": ".cookie-banner", "timeout": 5000 },
    { "action": "click", "selector": ".accept-btn", "delay": 300 }
  ]
}

Reference them in the config:

{
  "include": ["./steps/dismiss-banner.json"],
  "videos": { ... }
}

Multiple videos in one config

Define several videos in the videos map. Shared settings (viewport, theme, defaultDelay) are inherited from the top level.

Environment variables

Config values support $VAR and ${VAR} substitution from the environment.

Output formats

Set the output extension to control format: .mp4 (default), .gif, .webm.

{ "output": "demo.gif" }

Tips

  • Always set waitFor on a video to ensure the page is ready before steps run.
  • Use delay on individual steps to control pacing between actions.
  • Use --watch during development for automatic re-recording on config changes.
  • Use composite to iterate on theme/overlay settings without re-recording.
  • Use --verbose to debug step execution.
  • Use --dry-run to inspect the fully resolved config (includes, env vars, defaults).
  • Use zoom to scale up small UIs for readability in the recording.
  • Start with preview to verify steps work before committing to a full recording.

Troubleshooting

Browser Errors

Error: No inspectable targets
Error: Browser closed unexpectedly

Cause: Almost always a config issue, not a missing browser. webreel auto-downloads Chrome to ~/.webreel.

Fix:

  1. Run npx webreel record --verbose to see which step fails
  2. Check waitFor — if the selector doesn't exist on the page, the browser times out
  3. Run npx webreel install --force to re-download binaries if they're corrupted
  4. If using a custom CHROME_PATH, verify the binary exists and is executable

Missing Elements

Error: Element not found: "Submit Button"
Error: Timeout waiting for selector ".hero-section"

Cause: The element hasn't loaded yet, the selector is wrong, or the page layout changed.

Fix:

  1. Add a wait step before interacting: { "action": "wait", "selector": ".hero-section", "timeout": 10000 }
  2. Use npx webreel preview --verbose to watch the browser and see what's on screen
  3. Verify selectors with browser DevTools on the target URL
  4. If the page has lazy-loaded content, add a scroll step to trigger loading first

Timing Issues

Symptoms: Actions happen too fast, elements are clicked before animations complete, typed text is cut off.

Fix:

  • Add delay to individual steps: { "action": "click", "text": "Submit", "delay": 500 }
  • Set defaultDelay at the top level for consistent pacing across all steps
  • Use clickDwell to add a pause before clicks (makes cursor movement feel natural)
  • For typing, use charDelay to control speed: { "action": "type", "text": "hello", "charDelay": 80 }

Recording Produces Black/Empty Video

Cause: ffmpeg encoding issue or the page didn't render.

Fix:

  1. Run npx webreel record --frames to save raw JPEGs — check if frames are captured
  2. Verify the URL is accessible (not behind auth, VPN, or localhost that isn't running)
  3. Try a simpler config with just a pause step to isolate the issue

Reference files

  • steps-reference.md - detailed docs for all 12 step types
  • examples.md - annotated config examples for common use cases

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.