An offline-first field guide and local ledger for understanding multi-step Muse workflows without pretending to have telemetry you cannot see.
AI agents can search, read, click, upload, download, call tools, and modify data faster than a person can reconstruct what happened afterward. That makes it easy to confuse model tokens with network traffic, estimates with measurements, or a successful-looking result with a properly bounded action.
FedPromptly Muse Audit exists to make those boundaries visible. It gives you copy-ready prompts, practical lessons, and a device-local ledger for recording evidence one action at a time. It is independent of Meta, does not connect to a Meta account, and does not claim access to private Muse telemetry.
The demo shows the real PWA flow: open the audit mode, add an action to the local ledger, review the saved entry, and export the record. Nothing is sent to a server.
| Audit mode | Local ledger | Copy-ready workflow |
|---|---|---|
 |
 |
 |
Use these light-hearted images in release posts, social updates, or project discussions:
- Web/PWA: the self-contained static app in
site/with a device-locallocalStorageledger. - Skills Guide: five ready-to-install examples in
SKILLS.md, plus the source files inskills/. - Lessons and modules: a progressive curriculum from orientation through capstone projects in
lessons/andmodules/. - Desktop and Android shells: optional wrappers around the same canonical site under
desktop/andandroid/. - Privacy boundaries: guidance that keeps measured, reported, estimated, and unavailable values separate.
git clone https://github.com/Git-FED/MuseAi-Skills-Guide.git
cd MuseAi-Skills-Guide/site
python3 -m http.server 8080Open http://localhost:8080/. The main audit experience is at http://localhost:8080/usage-audit.html.
The PWA works as static HTML, including from file://. Ledger entries remain in the current browser or WebView until you explicitly export or clear them.
Start with the Skills Guide. Each example is a standalone Markdown skill that can be copied into a skills directory or adapted into a project instruction:
mkdir -p .skills
cp skills/audit-ledger/SKILL.md .skills/The examples cover audit-ledger tracking, privacy-boundary review, lesson planning, prompt review, and project scoping. They are intentionally small enough to inspect before installing.
See ROADMAP.md for the current plan and the public roadmap issue on GitHub.
Read SECURITY.md before installing an APK or desktop binary. The project is local-first: no account, backend, API key, analytics SDK, or automatic third-party loader is required. Explicit external links open only after user activation in the default system handler.
This project is available under the MIT License.