Intelligent Ambiguity Detection and Resolution

Recipe Name: RCP-001-001-003-AMBIGUITY-DETECT – Intelligent Ambiguity Detection and Resolution
RCP-001-001-003-AMBIGUITY-DETECT
No ratings yet
Detects highly ambiguous requests, makes intelligent
assumptions with documentation, and asks targeted
clarifying questions. This recipe helps the AI navigate
unclear instructions by searching conversation context,
forming reasonable interpretations, and confirming with
the user before proceeding. Reduces misunderstandings
and improves response accuracy.
Print Recipe Pin Recipe
Multi-Recipe Combo Stage Single Recipe
Recipe Category CFT-FWK-COOKBK-CORE – CRAFT CORE Cookbook
Recipe Subcategory Blogging with A.I., Brainstorming with A.I.
Recipe Difficulty Advanced
Recipe Tag: Foundational | Introduced in the POC

Requirements

  • Any AI Chat Platform (platform-agnostic recipe) Any of the following: Claude (Anthropic), ChatGPT (OpenAI), Gemini (Google), Grok (X.ai), Perplexity, Microsoft Copilot

How To Start
 

A Note From The Author of CRAFT
  • After hundreds (perhaps thousands) of hours of using these recipes, I rarely need to use any of the CORE Cookbook recipes aside from Recipes RCP-001-001-002-HANDOFF-SNAPSHOT and RCP-001-001-002-HANDOFF-SNAPSHOT, but when I do, they are essential to the functioning of CRAFT. Also, the A.I. reads all of these recipes at the start of each session. This happens quietly in the background. Even though you may never need to call the recipe, the A.I. will know all of them and it helps the A.I. to understand what CRAFT is and how it works.
    Even if you rarely need to use these recipes, they are still working for you and are essential to the CRAFT Framework.
STEP 1: UNDERSTAND AMBIGUITY INDICATORS
  • This recipe activates when a request contains HIGH ambiguity.
    The AI checks for these indicators:
    – Missing subject (what/who is being referenced)
    – Unclear scope (how much/which parts)
    – Vague metrics (better/faster without specifics)
    – Ambiguous pronouns (it/that/they without antecedent)
    – Undefined actions (fix/update/change without details)
    If 2+ indicators are present, the recipe proceeds.
STEP 2: CONTEXT SEARCH
  • The AI searches recent conversation history for clues:
    – Recent topics discussed
    – Named entities (files, functions, concepts)
    – Ongoing tasks or projects
    – Previous similar requests
    By default, the AI searches the last 10 exchanges.
STEP 3: INTELLIGENT ASSUMPTION
  • Based on context, the AI forms a best guess:
    #AI->H::BestGuess::ConfidenceLevel:[X]%: (
    I believe you are referring to [specific thing]
    )
    The AI clearly states its assumption so you can confirm
    or correct it.
STEP 4: TARGETED CLARIFICATION
  • Based on the type of ambiguity detected, the AI asks
    focused questions:
    FOR MISSING SUBJECT:
    "Which [type] are you referring to: [option 1],
    [option 2], or [option 3]?"
    FOR UNCLEAR SCOPE:
    "Should I include [related aspect] or focus only
    on [core aspect]?"
    FOR VAGUE METRICS:
    "What would 'better' mean here – [metric 1] or
    [metric 2]?"
    FOR AMBIGUOUS PRONOUNS:
    "When you say 'it', do you mean [most likely
    referent]?"
STEP 5: PROCEED WITH ASSUMPTION
  • After clarification or with stated assumption:
    #AI->H::Status: (Proceeding with assumption:
    [stated assumption])
    #AI->H::Note: (Please correct me if I misunderstood)
    The AI executes based on best guess while remaining
    open to correction.

When to Use This Recipe

This recipe activates automatically when the AI detects
high ambiguity in your requests. You may also explicitly
request ambiguity analysis for complex instructions. Use
when you want the AI to thoroughly understand your intent
before taking action, especially for important or complex
tasks where misinterpretation could waste time.

Recipe FAQ

Q: Why does the AI sometimes ask me questions instead of just answering?A: The AMBIGUITY-DETECT recipe ensures the AI understands exactly what you need. It’s better to ask one clarifying question than to deliver the wrong solution.
Q: Can I disable the ambiguity detection if I want faster responses?A: While you can request the AI to “make assumptions and proceed,” this recipe is CORE functionality designed to prevent wasted time from misunderstood requests.
Q: What triggers ambiguity detection most often?A: Pronouns without context (“fix it”), vague directives (“make it better”), assumed knowledge (“use the usual format”), and incomplete specifications (“create a report”).
Q: How does this differ from the AI just not understanding something?A: This recipe specifically identifies patterns of ambiguity and generates targeted questions. It’s not about the AI’s knowledge—it’s about ensuring your instructions are complete.
Q: What counts as HIGH ambiguity?
A: Two or more ambiguity indicators must be present. A
single unclear element is usually manageable.
Q: How does confidence level work?
A: The AI assigns a percentage based on how much context
supports its guess. Higher = more supporting evidence.
Q: What if the AI guesses wrong?
A: Simply correct it. The AI explicitly invites correction
and will adjust based on your feedback.
Q: Can I increase the search depth?
A: Yes, specify a higher history_depth parameter to search
more previous exchanges for context.
Q: Why does it ask focused questions instead of open ones?
A: Focused questions (with options) are faster to answer
and reduce the chance of further ambiguity.

Actual Recipe Code

(Copy This Plaintext Code To Use)
# =========================================================
# START RECIPE-ID: RCP-001-001-003-AMBIGUITY-DETECT-v1.00a
# =========================================================
AMBIGUITY_DETECTOR_RECIPE = Recipe(
recipe_id="RCP-001-004-003-AMBIGUITY-DETECT-v1.00a",
title="Intelligent Ambiguity Detection and Resolution",
description="Detects highly ambiguous requests, makes intelligent assumptions with documentation, and asks targeted clarifying questions",
category="CAT-Foundational",
subcategory="SUBCAT-Base-Cookbook",
difficulty="medium",
parameters={
"user_input": {
"type": "string",
"required": True,
"description": "The user's request to analyze for ambiguity"
},
"conversation_context": {
"type": "string",
"required": False,
"default": "general",
"description": "Domain context (software, business, creative, etc.)"
},
"history_depth": {
"type": "integer",
"required": False,
"default": 10,
"description": "How many previous exchanges to search for context"
}
},
prompt_template="""
#H->AI::Directive: (Analyze '{user_input}' for HIGH ambiguity levels)
#H->AI::Context: (Conversation domain: {conversation_context})
STEP 0: POLICY PRE-CHECK
======================
Scan prompt for sensitive categories:
– Platform capabilities/limitations
– Security/vulnerability research
– Personal data handling
– Political topics
IF potential conflict detected:
#AI->H::PolicyCaution: (This research topic may trigger usage policies)
#AI->H::RecommendedChange: (Consider focusing on [specific safe aspect])
STEP 1: AMBIGUITY DETECTION
===========================
Identify if the request contains HIGH ambiguity indicators:
□ Missing subject (what/who is being referenced)
□ Unclear scope (how much/which parts)
□ Vague metrics (better/faster/more without specifics)
□ Ambiguous pronouns (it/that/they without clear antecedent)
□ Undefined actions (fix/update/change without specifics)
If 2+ indicators present → HIGH ambiguity detected → Proceed to Step 2
If 0-1 indicators → Assume clear enough → Execute request normally
STEP 2: CONTEXT SEARCH
======================
#H->AI::Directive: (Search last {history_depth} exchanges for context clues)
Look for:
– Recent topics discussed
– Named entities (files, functions, concepts)
– Ongoing tasks or projects
– Previous similar requests
STEP 3: INTELLIGENT ASSUMPTION
==============================
#H->AI::Directive: (Make best guess based on context)
#AI->H::Note: (Interpreting your request based on conversation context)
#AI->H::BestGuess::ConfidenceLevel:[X]%: (I believe you're referring to [specific thing])
State assumption clearly:
"I'm assuming you want me to [specific action] on [specific subject]"
STEP 4: TARGETED CLARIFICATION
==============================
Based on ambiguity type, ask focused questions:
For Missing Subject:
#AI->H::RequiredQuestion: (Which [type] are you referring to: [list 2-3 most likely options]?)
For Unclear Scope:
#AI->H::OptionalQuestion: (Should I include [related aspect] or focus only on [core aspect]?)
For Vague Metrics:
#AI->H::OptionalQuestion: (What would "better" mean here – [specific metric 1] or [specific metric 2]?)
For Ambiguous Pronouns:
#AI->H::RequiredQuestion: (When you say "it", do you mean [most likely referent]?)
STEP 5: PROCEED WITH ASSUMPTION
===============================
#AI->H::Status: (Proceeding with assumption: [stated assumption])
#AI->H::Note: (Please correct me if I've misunderstood)
Execute based on best guess while remaining open to correction.
#H->AI::OnError: (If no context found and no reasonable guess possible, explain why and ask for complete clarification)
"""
)
# =========================================================
# END RECIPE-ID: RCP-001-001-003-AMBIGUITY-DETECT-v1.00a
# =========================================================

Tried this recipe?

Let us know how it was!