claude-skills / store-assets
CLAUDE.md0 starsChanged 30 days agoMITView on GitHub
Rules and commands bound to /store-assets's own tooling. Cross-cutting rules (lib-orchestrator.sh, skill directory shape) live in the root CLAUDE.md. Always render before validating. render.mjs's own pre-flight already fails before any render on a broken config, a missing source PNG, or a missing bezel/status-bar asset, but it does not check the previous run's output; validate.mjs is what checks the files actually on disk, so a stale or partial generated/ tree from an earlier round needs a fresh render first. Neither command…
- Installs packages
In this repository13 other files it gives its agents.
- audit/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- delegate/CLAUDE.mdCLAUDE.md
- plan-it/CLAUDE.mdCLAUDE.md
- screens/CLAUDE.mdCLAUDE.md
- ship/CLAUDE.mdCLAUDE.md
- auditSkill
- delegateSkill
- plan-itSkill
- readinessSkill
- screensSkill
- shipSkill
- store-assetsSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 25
Playwright fallback lives in
~/.cache/store-assets/npm: When the project has no own playwright/@playwright/test, render.mjs runs npm …
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md 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.
Files
/store-assets internals
Rules and commands bound to /store-assets's own tooling. Cross-cutting rules
(lib-orchestrator.sh, skill directory shape) live in the root CLAUDE.md.
Commands
| Command | Purpose |
node store-assets/bin/render.mjs --project <path> [--scene <id>] [--format <id>] [--locale <code>] [--background <hex>] [--out <dir>] | Renders stills. No filters = full run (every scene x locale x format in the project's config), plus generated/index.html (a static contact-sheet grid grouped by format/locale). --scene/--format/--locale narrow to one combination (used for spot checks). --background/--out: renders a uniform-palette variant into <dir> instead of the project's main generated/<format>/<locale>/ tree, without touching the config file (references/config-schema.md "CLI flags") |
node store-assets/bin/validate.mjs --project <path> | Checks every rendered file's exact dimensions, no alpha channel, sRGB, file size, per-format counts, and that every scene's text is reviewed: true with a matching reviewed_hash. Exit 0 only when everything passes |
Always render before validating. render.mjs's own pre-flight already fails before any render on a broken config, a missing source PNG, or a missing bezel/status-bar asset, but it does not check the previous run's output; validate.mjs is what checks the files actually on disk, so a stale or partial generated/ tree from an earlier round needs a fresh render first. Neither command deletes the project's own generated/ files itself: a schema change that drops or renumbers scenes (e.g. a scene-id rename) can leave stale files behind, remove them by hand before the next full render.
Rebuilding the bezel/status-bar cache on a new machine
None of ~/.cache/store-assets/bezels/, ~/.cache/store-assets/statusbar/, or the Android SDK's
~/Library/Android/sdk/skins/pixel_9_pro/{back,mask}.webp are downloaded or generated by the skill
itself; see references/store-specs.md "Devices" > "Asset provenance" for the exact DMG URL,
simctl status_bar override invocation, and Android SystemUI demo-mode steps. render.mjs's
pre-flight fails with the missing path before any render if the cache is not there yet.
Gotchas
- Playwright fallback lives in
~/.cache/store-assets/npm. When the project has no ownplaywright/@playwright/test,render.mjsrunsnpm install --prefix ~/.cache/store-assets/npm playwright@<pin>once (notnpx, which gives no importable module path), imports it from there, and setsPLAYWRIGHT_BROWSERS_PATH=~/.cache/store-assets/browsers, runninginstall chromiumonly when that dir has nochromium*entry. Nothing is installed into the project or the skill. bin/lib.mjsis the single source of truth forreviewedHash()andparseArgs(). Bothrender.mjs's pre-flight andvalidate.mjs's real gate call the exact same function with the exact same arguments (reviewedHash(text.headline, text.subline), no fallback); a scene with nosublinekey (a headline-onlycomboscene) hashessublineas JavaScript'sundefinedcoerced to a string, not empty string. Do not reintroduce a|| ''fallback at only one of the two call sites, they will disagree again.templates/scene.htmlnever hardcodes a machine path, brand value, or project copy. Device bezel/status-bar paths come fromreferences/store-specs.md's "Devices" JSON (extracted at render time byrender.mjs, not parsed by the browser); fonts/colors/logo/headlines/pills come from the project's own.store-assets/store-assets.jsonvia query params. Acomboscene's compositional constants (rotation angles, overlap offsets, margins) are the one exception: they are skill-level art-direction, not project config, and live as named constants inscene.html's own combo branch.ios-ipad-13is frameless.DEVICE_BY_FORMATmaps it to theipadDevices entry (frameless: true); pre-flight skips bezel/status-bar checks for it and instead requiressource_ipadper scene (front part forcombo, which renders as one single device). Rounded corner, width, top edge and shadow constants live in the Devices JSON (ipad, andipad-landscapeforios-ipad-13-landscape),scene.html's frameless branch only reads them fromdeviceSpec. Config overrides (format_options.<id>.width_pct/top_pct, sceneipad_*_pct, sceneformats) are validated inrender.mjs's pre-flight; defaults keep existing projects' output unchanged.- Milestone B (video,
templates/preview.html+bin/render-video.mjs) is not built. Nothing inSKILL.md's Ablauf beyond stills is implemented yet; see the plan's Step 10.
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
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.

