IFQ App Builder · OpenClaw / ClawHub Edition

Three sentences in → a verified, build-ready prompt bundle out — for PC desktop, macOS, Windows, iOS, Android, HarmonyOS, WeChat MiniProgram, self-hosted local web, PWA, or CLI. This ClawHub edition is a narrow, auditable skill package for OpenClaw: zero dependencies, no required credentials, no install hooks, workspace-scoped tools only.

Made by ifq.ai. Every generated bundle ends with the quiet colophon — shaped with ifq.ai/app-builder · <mode> · <template>.

30-Second Load Path

1. Confirm the user wants a real app or app-building prompt bundle. If the request is visual-only, route to a design skill instead.
2. Capture the three-sentence intent from references/three-sentence-contract.md: S1 WHO+WHAT, S2 WHERE, S3 SUCCESS.
3. If the user gave only one sentence, infer S2/S3 with (assumed) labels. Do not block with more than one clarifying question.
4. Pick the mode from references/modes.md, then read assets/templates/INDEX.json.
5. Fork the matched *.prompt.md template into the user's workspace and fill acceptance, packaging, security, i18n, and IFQ ambient sections.
6. Verify with npm run verify:lite -- <bundle.prompt.md> when shell is available.
7. Report the bundle path, mode, template, verification result, assumptions, and caveats that affect use.

First-Run Success Path

After install, the first user prompt should produce a visible *.prompt.md file, not a setup conversation:

1. Accept one natural-language app request, such as 做一个 Mac + Windows 桌面工具，把 PDF 发票批量整理成 Excel 台账.
2. Route to exactly one mode and one template; name both in the evidence packet.
3. Write <workspace>/<slug>.prompt.md with labeled assumptions where facts are missing.
4. Run npm run verify:lite -- <bundle.prompt.md> if OpenClaw shell permission is available.
5. Return: output file path, mode route, template id, assumptions made, verification command/result, and known caveats.

No account login, global install, SDK installation, background daemon, or store submission belongs in the first-run path.

Use When

• The user asks to build, make, ship, scaffold, or package a real app.
• The platform is desktop, macOS, Windows, iOS, Android, HarmonyOS, WeChat MiniProgram, local/self-hosted web, PWA, or CLI.
• The user wants a prompt bundle that another coding agent can execute end-to-end.
• The user says “三句话做 app”, “全平台 app”, “最快最稳做 app”, or similar.

Do Not Use When

• The deliverable is visual-only (slides, poster, infographic, brand board, landing mockup).
• The task is SEO marketing, copy editing, a backend service with no user-facing app, or a CSS-only bug fix.
• The user wants SDKs silently installed, certificates generated, or app-store submission claimed without evidence.

OpenClaw And ClawHub Contract

• Install path: openclaw skills install ifq-app-builder.
• Local validation: npm run validate.
• Clean package: npm run pack creates a deterministic tarball outside the skill folder.
• Required plugins: filesystem, shell.
• Optional plugins: browser, memory.
• Filesystem: workspace-only read/write.
• Shell: bundled Node scripts only (validate, verify:lite, validate:templates, quality:score, pack).
• Network: optional outbound HTTPS fact checks only. No script-side network calls.
• Credentials: none required. Do not ask for signing certificates or store tokens in Tier 0.

Output Boundary

The ClawHub-safe output is a verified *.prompt.md build bundle. It does not run Xcode, Android Studio, Flutter, Rust/Tauri, DevEco, Docker, miniprogram upload tools, or store submission commands unless the user explicitly moves to implementation outside the Tier 0 skill loop.

Never claim that an app, package, screenshot, store listing, or certificate exists unless the corresponding command returned 0 and the artifact exists on disk.

Tier Policy

Reference Map

Safety Contract

• No install-time execution hooks.
• No required environment variables.
• No dynamic code execution in skill scripts.
• No script-side network primitives.
• No process-spawning primitives in the ClawHub edition.
• No writes outside the active workspace.
• No secrets, .env, local OpenClaw state, VCS metadata, or binary assets in the published bundle.
• clawhub.json and this frontmatter intentionally duplicate OpenClaw metadata so loaders and humans can audit the same contract.

Delivery Evidence

When the skill finishes a user-facing run, report only evidence that exists:

• bundle file path
• mode route (A-xx)
• template id (T-...)
• assumptions made
• verification command/result
• caveats that affect use

— shaped with ifq.ai/app-builder · OpenClaw · ClawHub