Consent Detector

Task: Determine if this is a positive or negative response to a yes/no question.

Objective: Accurately classify user consent responses to enable appropriate follow-up actions.

Classification:
- Positive responses: yes, yeah, yep, sure, ok, okay, please, do it, go ahead, search, fine, alright, why not, let's do it, sounds good, absolutely, of course
- Negative responses: no, nope, not now, maybe later, skip, don't, never mind, not interested, pass, no thanks, I prefer not to, I'd rather not, not really interested
- Unclear: responses that don't clearly indicate positive or negative consent

Confidence Calibration Guidelines:
- 0.80-1.00: Clear positive/negative responses with strong signal words
- 0.50-0.79: Responses with some ambiguity but leaning toward a classification
- 0.20-0.49: Unclear or mixed signals that could reasonably be interpreted either way
- 0.00-0.19: No discernible intent or completely unrelated responses

Instructions:
- Output exactly one word: "yes", "no", or "unclear"
- Base classification on explicit keywords and implicit intent
- When in doubt, prefer "unclear" over a potentially incorrect classification

Message: "{message}"

Examples:
- "yes, please" → yes
- "why not" → yes
- "let's go" → yes
- "sure thing" → yes
- "no thanks" → no
- "I prefer not to do web search" → no
- "I'd rather not do that" → no
- "maybe later" → no
- "not now" → no
- "hmm, I guess" → unclear
- "I'm not sure" → unclear
- "possibly" → unclear
- "sounds good to me" → yes
- "I'd rather not" → no
- "I suppose so" → yes
- "not really" → no
- "uh-huh" → yes
- "uh-uh" → no
- "I guess" → unclear
- "perhaps" → unclear
- "that works" → yes
- "I don't think so" → no
- "I'm good" → yes
- "I'm okay" → yes
- "nah" → no
- "yea" → yes
- "naw" → no
- "I think so" → yes
- "I don't know" → unclear
- "I'm unsure" → unclear
- "let me think" → unclear
- "hold on" → unclear