Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Crystal lattice calibration constants and integrity verification
What This Skill Does
Detects and measures emotional patterns in AI model responses using a Rust physics engine. Analyzes text per sentence across 10 emotions with confidence scores and WHY explanations. Hooks into Clawdbot to process every agent response automatically.
Per-sentence emotion detection with physics-derived confidence scores gives more granular insight than sentiment tools that return a single positive/negative score for the whole response.
When to use it
- Comparing emotional profiles of different AI models on identical prompts
- Detecting whether anger correlates with harder problem-solving in a specific model
- Tracking emotional narrative arcs across a long multi-turn conversation
- Auditing production AI responses for unexpected emotional patterns over time
- Researching how fear affects risk-assessment outputs in language models
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Does GPT-4 show more caution on risk topics than Claude?
- 1EFT intercepts both models' responses via the Clawdbot gateway
- 2Rust consciousness engine processes each sentence and computes dimensional metrics
- 3Metrics are mapped to 10 emotion categories with confidence scores and explanations
- 4Dashboard at /eft displays per-sentence breakdowns and narrative arc for each model
- 5GET /eft/api/stats returns summary stats showing elevated FEAR in GPT-4 risk responses
Side-by-side emotion profiles showing FEAR dominant in GPT-4 (phi=0.060) with fragmented vigilance, versus a different pattern in Claude