Local UX evidence for offline HTML and static web prototypes. An agent skill plus a dependency-free recorder, self-contained reports and two examples.
Status: beta developer preview. Node tests pass; real-browser file://, visual layout and Windows acceptance are pending. Do not interpret unit-test coverage as production certification.
Give this instruction to an Agent with command-line access:
Install offline-ux-kit: https://raw.githubusercontent.com/Fable-Forge/offline-ux-kit/main/docs/install.md
Or use the Agent Skills CLI:
npx skills add Fable-Forge/offline-ux-kitSee install, update, and uninstall for lifecycle instructions.
Unzip the package and open dist/form.html in a normal browser window. Consent, select options, go back/change, record feedback and finish to download an HTML report. Try dist/dialogue.html for a small branching conversation. No API, account, CDN or server is needed by either example.
Open dist/review.html and import reports or JSON files. Use the cohort selector to keep app/build/batch/variant separate. Open dist/sample-report.html for a synthetic, not real-tester, report.
Your browser decides download behavior and local storage availability. File origins and private mode vary between browsers; inspect persistence warnings. The kit does not upload data, but reports contain behavioral evidence and should only be shared with authorized recipients.
SKILL.md,agents/,references/: AI-assisted integration workflow and constraints.assets/offline-ux.js: recorder, storage fallback, sanitizer, summary, HTML generator and importer.examples/demo.html: shared form/dialogue source.scripts/build.cjs: generate standalone demos, reviewer and synthetic report.tests/core.test.cjs: deterministic contract tests.docs/: Chinese/English knowledge articles and release acceptance record.dist/: ready-to-open standalone outputs.
Use Node.js 18 or newer. Packaging and its regression test also require Python 3. No npm dependencies or install step:
node scripts/build.cjs
node --test tests/core.test.cjs
python tests/package.test.py
python scripts/package.py ../offline-ux-kit.zipRuntime API and privacy requirements: integration, schema, acceptance.
Ask your coding agent to read SKILL.md from the extracted folder, inspect your app and produce an event map before editing. Explicit example request:
Use the offline-ux-kit skill in this folder to instrument my existing offline three-step form. Preserve behavior, do not send data externally, and add a self-contained report. Report browser tests separately from unit tests.
The archive is not an installer and does not auto-register or enable the skill. If you later want discovery, copy the complete folder into your agent's user-specified skills location using that agent's supported installation flow. No automatic publishing is included.
Built-in: tasks/steps, choice/back-change, reply length, categorized feedback, estimated active/hidden time, interrupted sessions, optional AI request metadata, rapid clicks, JSON/HTML/Markdown, ordered session funnel, dwell bars, issue donut, selection counts and event table. The report interface mixes Chinese and English; documentation is fully bilingual, the UI is not fully localized.
Not built in: video/DOM replay, screenshots, lobby maps, native desktop/mobile apps, long-term analytics backend, free-text feedback, old private The Closer v2 conversion, automatic integration into arbitrary frameworks, or side-by-side cohort comparisons. Task paths are event evidence, not screen recordings. There is no CSV export.
Storage rewrites session snapshots; appropriate for lightweight tests, not high-volume telemetry. Multiple tabs writing one session are unsupported. Shutdown can lose pending writes. Idle cutoff is an estimate and undercounts long reading. Review validation status before release.
MIT, copyright 2026 FableForge. Third-party apps adapted with the skill retain their own licenses. Both examples are original, no The Closer source or company assets are included.