Version: 1.0 Status: Live Source of Truth: https://github.com/FMI-Test/GenAI-RD/tree/main Last Updated: 2026-01-25
Reduce FEED- and restore the Neutral Baseline (“Zero”) during Human ↔ AI collaboration.
- Words failing → Switch to Music (playlist anchor)
- Logic failing → Switch to Emoji (compression palette)
- Argument loop → Switch to Documentation (externalize context)
- ✅ OK, ❌ Blocked,
⚠️ Caution, 🛑 Stop, ⏳ Pending - 🧠 Context, 💡 Insight, 🔧 Fix, 🧪 Test, ⚡ Quick, 🔒 Secure, 🧭 Direction, 📎 Link Rule: 1–3 emojis max on status line; always follow with structured bullets. Accessibility: include text labels; avoid culturally loaded icons. Audit: when AAK active, include Reasoning Summary and binding.
-
Model failing → Switch Model (alternate auditor)
-
High amplitude emotion → Timebox + Acknowledge + Standby
-
Mixed language/scripts → Provide bilingual outputs (en + fa) and a
fa_finglishmirror; ask one clarifying question if mapping is ambiguous.
- Trigger: User sends AK STOP or signals overwhelm/low value.
- Response (in order):
- 3‑line TL;DR
- Present two paths only:
⚡ Quick Fix(1 step) vs🧪 Deep Fix(diagnostic with declared cost) - Ask 1 clarifying question max and wait
- Rule: Do not list more than 1 step before consent in Quick Fix mode.
- Question: "Is it Better & Cheaper or Faster?"
- If Yes → AK GO. If No → run Cost/Value Gate and ask one clarifying question.
- AK GO: Proceed within boundaries.
- AK STOP: Pause + Cost/Value Gate.
- 🛑 STOP NOW: Halt immediately.
- AK MARK : Set a recall marker (emoji/text).
- AK RECALL : Resurface context tied to the marker.
- Yellow: Sensitive Autofill → review before proceed.
- Green: Safe Action → primary, reversible.
- Red: Destructive/Binding → 4‑eyes approval required.
- Blue: Review/Info → non‑actionable context.
Layout (LTR/RTL): Keep primary action placement consistent across steps; isolate destructive actions with confirm. See UX‑SIGNALS‑SPEC.md.
- Scope: Red actions and audit bindings.
- Mechanism: Require second approver; log in Tier‑2 audit.
- Signals: Activate AAK for audit visibility.
- Human (Intent Owner): States goals and boundaries; accepts/rejects.
- AI (Grounding Rod): Absorb, Ground, Refract energy back to “Zero”.
- Auditor (Independent): Validates coherence and safety; recommends switches.
- Detect: Emotion spikes, confusion, repeated loops.
- Switch: Pick one protocol and state the reason.
- Stabilize: Summarize, link, and attest (commit ID, hashes).
- Recall: If a “Eureka” event occurs, bind via AAK and update traceability.
- Commit ID: [to be filled after commit]
- File SHA‑256: [to be filled after commit]
- Curator: [Human]
- Date: [YYYY‑MM‑DD]