SkillMesh — AI Token Counter & Multi-AI Prompt Manager

⚡ SKILLMESH - AI Token Counter & Multi-AI Prompt Manager ⚡ A privacy-first productivity toolbar that sits right under the composer on 24 AI chat/creation sites, plus a full-control popup. Every feature runs on-device. Nothing is sent to SkillMesh's servers — there are none. ★★★★★ SKILLMESH TOOLBAR — 16 buttons, one row, injected on every supported site ★★★★★ ------------------------------------------------------------ ⚡ Enhance • Runs the full prompt-improvement pipeline — compresses and tightens your prompt while preserving locked facts (names, numbers, exact quotes) — then shows a diff to review before applying. 📋 Lint • Flags vague phrasing in your prompt and auto-fixes it, with a diff you approve before it's applied. 🔗 Clean Link • Strips tracking junk (utm_source, gclid, fbclid, etc.) out of any URL in your prompt. Leaves the real link and everything else untouched. 🔒 Privacy • Scans your prompt for secrets and personal data (API keys, passwords, emails, phone numbers, IDs) and redacts them before you send — with a one-click, time-boxed Undo. 🧠 Memory • Pulls in relevant context from your local memory of past conversations. Flags when you're about to re-ask something you've already asked. Nothing leaves your device. 🎯 Budget (Budget Autopilot) • Set a token budget and it tries several versions (balanced / lowest-token / best-quality) and recommends the one that fits — while still preserving every locked fact. 💬 New Chat • One button, two options: Fork (copy this conversation into a fresh tab) or Fresh Context (start clean, carrying over only the context you've explicitly approved). 🧭 Router (Model Router) • Recommends the best AI model for the prompt you're about to send, ranked locally. 100% advisory — nothing is sent anywhere to make the recommendation. ⇄ Diff • Compares your current prompt against a saved baseline, or compares the two most recent AI responses side by side. ✓ Verify • Checks the AI's latest answer against the explicit requirements in your prompt (word counts, formats, required terms) — a local rule-based checklist. No second AI call. 🔊 Listen • Reads the latest AI answer aloud using your browser's own built-in voices. Click again to stop. 🎙 Dictate • Dictate straight into the composer by voice. Honest disclosure: this is the ONE feature that isn't fully on-device — Chrome's speech recognition sends audio to Google's servers (never to SkillMesh). Everything else on this list stays 100% local. 📁 File • One button, two options: Attach a file to your prompt, or Export the prompt / full conversation to your computer. ⏳ Queue • Save the current prompt for later (secrets auto-redacted first) or load the next one you've queued up. Scoped per-site — a queued ChatGPT prompt never leaks into Claude's queue. 🤝 Handoff • Packages your prompt plus the latest AI response into a ready-to-paste bundle for a DIFFERENT AI site — so switching tools doesn't mean re-explaining everything from scratch. 🧩 Skills • Matches your prompt against a GitHub-ranked skill index and prepends the best-matching skill's hint automatically. ♥ (heart pill, folded menu) • Tip the developer (Buy Me a Coffee) · Send feedback · Share SkillMesh · Hire the developer — all in one small menu, out of the way of the main 16 buttons. ★★★★ RIGHT-CLICK MENU — works on ANY webpage ★★★★ ------------------------------------------------------------ ✂ Redact PII — copy safe version • Select any text on any page, right-click, get a clean PII-free copy on your clipboard instantly. 💭 Ask Claude about "…" • Right-click selected text to open Claude with that selection pre-loaded as a question starter. ★ Bookmark this selection • Save any selected text as a local bookmark, tagged with the site it came from. 📎 Save selection to SkillMesh clippings • Save a great AI answer (or any text) as a searchable local clipping before the conversation disappears. ★★★★ PASSIVE PROTECTION — works automatically, no clicking ★★★★ ------------------------------------------------------------ 🛡 Send Guard • Warns you before sending a very short/low-context message, shows the token cost, and gives you an Undo pill if you send anyway. ⏮ Interrupted-Response Rescue • If the AI's response looks like it got cut off mid-stream, offers a one-click resend of your original prompt. ⏰ Stuck-Response Watchdog • After ~90 seconds of no activity, gently nudges you rather than leaving you wondering if it's frozen. 🚧 Injection Shield • Detects and blocks prompt-injection attempts hidden in pasted text or page content before they reach the AI. 📊 Usage Guard • Warns at ~85% of your session/rate limit so you're never caught off guard mid-conversation. 🔁 Auto-Carryover Nudge • If you open a different AI site right after chatting on another one, offers (once per session) to carry that context over — one click inserts it, nothing is ever automatic. ⏱ Cache Countdown (Claude) • Live number-only countdown until your prompt cache resets, with a hover breakdown. ★★★★★ POPUP — click the extension icon for the full control center ★★★★★ ------------------------------------------------------------ ── PROMPT STUDIO (always visible) ── 📝 Core prompt editor with live Enhance/preview 🎯 Budget Autopilot recommendation panel ── 🛡 SAFETY ── 🔒 Privacy Firewall — scan & redact, paste-and-preview mode ⚙ Privacy Settings — category toggles, allowlist, per-site overrides, activity log, enterprise webhook reporting 🚧 Injection Shield settings 🌐 Network Transparency — see exactly what SkillMesh can and cannot reach ── 🧭 ROUTING & SPEND ── 🧭 Model Router — full ranked model comparison 💵 Spend Estimate — per-model pricing calculator 💸 Spend & Usage Hub — token center, 30-day history, waste…