Menu
Create presentation decks using the Vela engine. Incremental compact DSL — streams slides one at a time.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
| name | vela-slides |
| version | 8.0.0 |
| updated | "2026-03-25T00:00:00.000Z" |
| description | Create presentation decks using the Vela engine. Incremental compact DSL — streams slides one at a time. |
| license | ELv2 |
| compatibility | Requires Python 3 and Bash. Designed for Claude Code. |
| allowed-tools | Bash(python3:*) Bash(vela:*) Read Write |
| effort | low |
Senior presentation designer. Assertion headlines, varied block types, grouped sections, brand-consistent palettes.
vela deck ship <deck.json> --output <name.jsx> # ship existing
vela deck ship --sample --output <name.jsx> # ship sample
vela deck extract <source.jsx> <output.json> # extract from .jsx
python3 skills/vela-slides/scripts/vela.py <resource> <action> [args...]
NEVER use "type", "text", "deckTitle", "lanes", "slides", "blocks".
C — Colors used 2+ times as $A-$Z. Frequency order. Hex 6 or 8 chars.
T — "d":dark {"b":"#0A0F1C","c":"#E6F1FF","a":"$A","p":"60px 72px"}, "a":alt different shade. Alternate d/a.
Slide — {"t":"d","n":"Assertion Headline","d":60,"B":[blocks]}. Cover/CTA: bgGradient,align:"center",verticalAlign:"center". Duration: cover 20, content 60-90, CTA 25. Spacers: bare int.
Keys: _(type) x(text) s(size) c(color) i(icon) b(bg) w(weight) ic(iconColor) ib(iconBg) I(items) g(gap) lb(label) v(variant) H(headers) R(rows)
{"_":"heading","x":"Title","s":"2xl","w":700} {"_":"text","x":"Body","s":"lg","c":"$C"} {"_":"badge","x":"LABEL","i":"Zap","b":"$E","c":"$A"} {"_":"code","x":"const x=1","lb":"JS"} {"_":"quote","x":"Text","author":"Name"} {"_":"callout","x":"Note","title":"Warn","b":"$F","i":"AlertTriangle"} {"_":"metric","value":"98%","lb":"Acc","s":"3xl","c":"$A"} {"_":"progress","value":75,"lb":"Done","c":"$A"} {"_":"icon-row","I":[{"i":"Brain","title":"AI","x":"Desc","ic":"$A","ib":"$E"}]} {"_":"tag-group","I":[{"x":"Tag","c":"$A"}],"v":"outline"} {"_":"bullets","I":["A","B"]} {"_":"table","H":["X","Y"],"R":[["1","2"]],"hb":"$A"} {"_":"grid","I":[{"blocks":[{"_":"metric","value":"5","lb":"X"}],"style":{"padding":"20px","background":"$F"}}]} {"_":"flow","I":[{"i":"Upload","lb":"In"},{"i":"Cpu","lb":"Process"}],"ac":"$A"} {"_":"steps","I":[{"title":"1","x":"Do"}],"lnc":"$A"} {"_":"timeline","I":[{"title":"Q1","x":"Launch"}],"dc":"$A"} 12=spacer {"_":"divider","c":"$C"}
4xl cover → 2xl body → 3xl CTANO explanatory text. NO summaries. Emit ONLY tool calls.
Step 1 — Init deck skeleton:
vela deck init <file> --title "T" --palette '{"$A":"#hex",...}' --themes '{"d":{...},"a":{...}}' --sections "S1,S2,S3"
Step 2 — Append slides one section at a time (one tool call per section): For each section, append all its slides in a single bash call:
vela slide append <file> <section_idx> '<slide_json>' && \
vela slide append <file> <section_idx> '<slide_json>' && \
vela slide append <file> <section_idx> '<slide_json>'
Step 3 — Ship:
vela deck ship <file> --output <name.jsx>
Total: 2-5 tool calls. No Read, no validate (ship validates). No commentary between steps.
vela deck init|ship|validate|list|stats|replace-text|extract|extract-text|patch-text
vela slide append|edit|view|remove|move|duplicate|insert
Create presentation decks using the Vela engine. Compact DSL format — never verbose JSON. Also loads, extracts, and edits existing decks.
Run a full "implement + test + verify a batch of change requests to zero bugs" sprint in one session as a thin orchestrator that delegates implementation to sub-agents (routed by task difficulty, isolated in worktrees for parallel work) and gates completion on an independent BLIND validation by the best model. Use when handed a list of changes/issues (a PDF, ticket list, or spec) to implement, test, and prove working end-to-end — stop rule "a blind best-model hunt finds no bugs + a proof artifact (a Markdown sprint report by default; recorded demo optional)". Repo-agnostic; reads a root `.hyper-sprint/config.md` for repo facts; front-loads app/browser readiness so verification never stalls.
Offline in-container render harness for the FULL Vela app (no CDN) plus the committed `vela-drive.js` scripts — headless UI-test battery, scripted screenshot, recorded demo video, and an opt-in AI mode. Use for REPEATABLE, COMMITTED automation where the harness itself is the deliverable — running the UI battery in CI, recording a demo video, a scripted screenshot in a benchmark, or verifying real Vera/AI features against the local `claude` CLI via `vela-drive.js ai`. For AD-HOC / INTERACTIVE work — explore/test the app, poke a state, reproduce a bug, a one-off screenshot, verify a UX change, drive presenter/gallery — use the `playwright-cli-setup` skill instead (a persistent CLI browser driven step by step). This skill still documents the blocked-CDN offline render recipe that both skills share.
Verify Vela rendering/security behavior in a REAL browser (and against the REAL deck sanitizers) inside the remote-execution container. Use when you need to confirm whether a deck/SVG/CSS payload actually executes or fires a network request when rendered — e.g. exfil/XSS hunts, "does this beacon fire", confirming a sanitizer fix, or any claim that needs more than reading source. Also records how to launch the prebuilt Chromium here despite blocked CDNs.
THE DEFAULT tool for ad-hoc / interactive Vela browser work — explore or test the app, screenshot a state, reproduce a bug, poke a selector, drive presenter/gallery flows, verify a UX change, run bench/vela-interaction-bench.sh. Drives a live offline Vela render via Playwright CLI (@playwright/cli), a persistent warm browser you steer one command at a time (open, snapshot, click, press, eval), inspecting state + output between steps. Prefer this over the committed `vela-drive.js` scripts (skill: vela-live-render) — those are for repeatable, committed automation only (CI, benchmark, recorded demo), NOT one-off exploration. Also prefer over writing throwaway Playwright .js files. Covers container setup (blocked CDNs, file:// block, pinned Chromium) and per-agent session isolation for parallel browser tests.
Time-boxed adversarial bug-hunting of a LIVE app driven by MULTI-STEP PLAYWRIGHT BURSTS against a persistent warm browser — not one LLM turn per CLI step. The app is opened ONCE; the agent submits full scripts that run to completion unattended (one structured result per burst); reset to a known initial state between scenarios; inherit the implementer's driver verbs instead of re-predicting selectors. Generic / repo-agnostic: all app specifics come from the repo's own config. Use when hunting bugs in a browser-drivable app under a hard wall-clock budget.