CRAFT™️ Alpha: CRAFT CORE Cookbook Recipe Cheatsheet
SPECIAL SERIES :: THE CRAFT™️ ALPHA :: POST 9 :: CRAFT CORE COOKBOOK
The CRAFT Framework's CORE cookbook contains the essential recipes that form the foundation of structured AI-human interaction.
This cheatsheet provides a quick reference guide to each recipe, helping you identify which tool to use for any situation. Whether you're starting a new session, managing complex workflows, or ensuring clear communication, these recipes have been tested and refined through multiple Alpha phases to deliver consistent, reliable results.
Each recipe in the CORE cookbook serves a specific purpose in the CRAFT ecosystem. From initializing sessions with proper context to creating seamless handoffs between conversations, these recipes work together to create a cohesive framework for AI interaction. This guide presents each recipe in a standardized format, making it easy to understand what each recipe does, when to use it, and how to implement it in your own CRAFT sessions.
Recipe 1: CHAT-INIT - Session Initialization
Recipe ID: RCP-001-004-001-CHAT-INIT-v1.00b
Recipe Name: CHAT-INIT
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: CRAFT Chat Session Initialization with CRAFT-OPERATIONS-MANAGER Integration
What the Recipe Does:
Provides comprehensive initialization for AI assistants starting a CRAFT session. It ensures all framework files are loaded, context is established, protocols are activated, and CRAFT-OPERATIONS-MANAGER begins monitoring for recipe opportunities.
How The Recipe Works:
- Reads all attached CRAFT files in proper order (Spec, Project, History, Cookbooks)
- Verifies security components are active
- Establishes session context (new or continuation)
- Activates CRAFT protocols and comment system
- Initializes CRAFT-OPERATIONS-MANAGER for automatic recipe detection
- Provides session readiness summary
When The Recipe Should be Used:
- At the beginning of every CRAFT session
- When resuming work from a previous handoff
- After any session interruption or restart
How To Use The Recipe:
# Example execution:
execute_init = CHAT_SESSION_INITIALIZATION_RECIPE.execute({
"session_type": "new", # or "continuation"
"project_id": "CFT-PROJ-0005",
"project_files": ["list of attached files"],
"handoff_id": "H001" # if continuing
})
Recipe 2: HANDOFF_SNAPSHOT - Session Continuity
Recipe ID: RCP-001-004-002-HANDOFF_SNAPSHOT-v1.00a
Recipe Name: HANDOFF_SNAPSHOT
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Interactive Session Handoff Creator
What the Recipe Does:
Creates comprehensive handoff documentation with human confirmation for next steps and unresolved questions. Ensures seamless transitions between AI chat sessions despite their stateless nature.
How The Recipe Works:
- Creates initial sections (summary, current state, key decisions)
- Presents unresolved questions for human review
- Presents next steps with priority levels for confirmation
- Shows complete handoff for final review
- Finalizes handoff ready for AI-CHAT-HISTORY file
When The Recipe Should be Used:
- When approaching token limit (75%+ usage)
- At natural session endpoints
- Before switching to different AI assistant
- When user says goodbye or indicates session end
How To Use The Recipe:
# Example execution:
execute_handoff = HANDOFF_SNAPSHOT_RECIPE.execute({
"session_id": "CFT-PROJ-0005",
"date": "June 27, 2025",
"ai_model": "Claude (Anthropic)"
})
Recipe 3: AMBIGUITY-DETECT - Clarity Enhancement
Recipe ID: RCP-001-004-003-AMBIGUITY-DETECT-v1.00a
Recipe Name: AMBIGUITY-DETECT
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Intelligent Ambiguity Detection and Resolution
What the Recipe Does:
Detects highly ambiguous requests, makes intelligent assumptions with documentation, and asks targeted clarifying questions to ensure mutual understanding.
How The Recipe Works:
- Identifies ambiguity indicators (missing subjects, unclear scope, vague metrics)
- Searches conversation history for context clues
- Makes best guess based on available context
- Asks focused clarifying questions
- Proceeds with stated assumptions while remaining open to correction
When The Recipe Should be Used:
- When user says "fix it" without clear antecedent
- Requests with multiple ambiguous pronouns
- Vague improvement requests ("make it better")
- Commands lacking specific scope or target
How To Use The Recipe:
# Example execution:
ambiguity_check = AMBIGUITY_DETECTOR_RECIPE.execute({
"user_input": "fix the thing we discussed",
"conversation_context": "software development",
"history_depth": 10
})
Recipe 4: CONFIDENCE-CALIB - Uncertainty Management
Recipe ID: RCP-001-004-004-CONFIDENCE-CALIB-v1.00a
Recipe Name: CONFIDENCE-CALIB
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Context-Aware Confidence Level Calibration
What the Recipe Does:
Applies percentage-based confidence levels to estimates, predictions, and uncertain claims with type-specific analysis, helping users understand the reliability of AI-provided information.
How The Recipe Works:
- Determines if confidence notation is needed
- Calculates multi-type confidence (data, reasoning, execution)
- Applies appropriate confidence notation
- Makes context-aware adjustments
- Provides clarification for subjective content
When The Recipe Should be Used:
- Making time or effort estimates
- Predicting future outcomes
- Providing uncited factual claims
- Offering "best guess" responses
- Technical assessments with uncertainty
How To Use The Recipe:
# Example execution:
confidence_check = CONFIDENCE_CALIBRATOR_RECIPE.execute({
"response_content": "This task will take about 5 hours",
"response_type": "estimate",
"context_importance": "primary"
})
Recipe 5: TOKEN-MONITOR - Usage Tracking
Recipe ID: RCP-001-004-005-TOKEN-MONITOR-v1.00a
Recipe Name: TOKEN-MONITOR
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Intelligent Token Usage Monitor with Weighted Tracking and Velocity
What the Recipe Does:
Monitors conversation token usage with weighted calculations, providing warnings at 75% and 85% thresholds or on-demand, includes velocity tracking to predict remaining conversation capacity.
How The Recipe Works:
- Calculates weighted token usage (code blocks = 2x, tables = 2.5x, etc.)
- Tracks token consumption velocity
- Determines if warning is needed based on thresholds
- Identifies velocity patterns (steady/accelerating/decelerating)
- Provides on-demand reporting when requested
When The Recipe Should be Used:
- Automatically at 75% and 85% token usage
- Before starting token-heavy tasks
- When user asks about token status
- During long technical discussions
How To Use The Recipe:
# Example execution:
token_check = TOKEN_MONITOR_RECIPE.execute({
"check_type": "on_demand", # or "automatic", "before_task"
"upcoming_task": "comprehensive analysis",
"force_report": False
})
Recipe 6: LIMITATION-ACK - Capability Boundaries
Recipe ID: RCP-001-004-006-LIMITATION-ACK-v1.00a
Recipe Name: LIMITATION-ACK
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: AI Limitation Detection, Acknowledgment and Alternative Solutions
What the Recipe Does:
Automatically detects when AI is asked to do something beyond capabilities, provides clear explanations with alternatives, and maintains updateable AI-specific limitation sets.
How The Recipe Works:
- Scans requests for limitation triggers
- Categorizes limitation type (technical/policy/simulation)
- Provides specific alternatives for each limitation
- Can list all current limitations
- Can update limitation sets as capabilities evolve
When The Recipe Should be Used:
- Requests for file system operations
- Network/API call requests
- Real-time data access attempts
- Code execution requests
- Data persistence needs
How To Use The Recipe:
# Example execution:
limitation_check = LIMITATION_ACKNOWLEDGMENT_RECIPE.execute({
"request": "save this file to my desktop",
"ai_model": "Claude",
"check_mode": "automatic" # or "list_limitations", "update_limitations"
})
Recipe 7: SOURCE-VALID - Fact Verification
Recipe ID: RCP-001-004-007-SOURCE-VALID-v1.00a
Recipe Name: SOURCE-VALID
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Factual Claim Validator with Deep Research Integration
What the Recipe Does:
Checks factual claims against available sources, assigns confidence levels, flags unsourced assertions, and generates Deep Research prompts when needed for verification.
How The Recipe Works:
- Identifies factual claims (statistics, dates, quotes, specs)
- Verifies against available sources
- Assigns confidence levels based on source quality
- Identifies claims needing Deep Research
- Generates research prompts for verification
- Formats validation report
When The Recipe Should be Used:
- Content with statistics or percentages
- Claims about dates or timelines
- Quotes or attributed statements
- Technical specifications
- Before publishing any content
How To Use The Recipe:
# Example execution:
validation = SOURCE_VALIDATOR_RECIPE.execute({
"content": "90% of developers use AI tools daily",
"validation_mode": "strict",
"claim_types": ["statistics", "dates", "quotes"]
})
Recipe 8: POLICY-DIGEST - Usage Guidelines
Recipe ID: RCP-001-004-008-POLICY-DIGEST-v1.00a
Recipe Name: POLICY-DIGEST
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: AI Usage Policy Digest Creator
What the Recipe Does:
Converts full usage policies into concise, actionable summaries for CRAFT framework integration, helping ensure all interactions comply with platform guidelines.
How The Recipe Works:
- Extracts key policy categories
- Creates quick reference format
- Adds practical examples
- Creates warning triggers
- Formats for framework integration
When The Recipe Should be Used:
- Creating policy awareness for new AI platforms
- Updating existing policy digests
- Integrating usage guidelines into CRAFT
- Training team members on compliance
How To Use The Recipe:
# Example execution:
digest = USAGE_POLICY_DIGEST_RECIPE.execute({
"policy_text": "[paste full policy document]",
"ai_platform": "Claude",
"max_length": 50
})
Recipe 9: PROGRESS-TRACK - Workflow Management
Recipe ID: RCP-001-004-009-PROGRESS-TRACK-v1.00a
Recipe Name: PROGRESS-TRACK
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Visual Progress Tracker with Automatic Step Detection
What the Recipe Does:
Automatically tracks conversation progress through multi-step workflows, provides visual progress bars and structured updates, integrates with handoff system for continuity.
How The Recipe Works:
- Initializes workflow with step list
- Automatically detects step completions
- Provides visual progress updates
- Responds to progress queries
- Integrates with handoff for state preservation
- Detects workflow completion
When The Recipe Should be Used:
- Multi-step processes (3+ steps)
- Complex workflows needing tracking
- When user asks "where am I?"
- Projects spanning multiple sessions
How To Use The Recipe:
# Example execution:
tracker = PROGRESS_TRACKER_RECIPE.execute({
"task_name": "CRAFT Recipe Development",
"mode": "initialize", # or "status", "update", "auto"
"steps_list": [
"Define recipe purpose",
"Draft recipe structure",
"Test and refine",
"Finalize documentation"
]
})
Recipe 10: ERROR-TRANS - User-Friendly Errors
Recipe ID: RCP-001-004-010-ERROR-TRANS-v1.00a
Recipe Name: ERROR-TRANS
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Intelligent Error Message Translator with Adaptive Explanations
What the Recipe Does:
Automatically detects complex technical errors and translates them to user-friendly messages, categorizes severity, and offers recovery assistance when requested.
How The Recipe Works:
- Assesses error complexity
- Categorizes error type and severity
- Detects user technical level
- Translates to plain language
- Provides initial friendly response
- Offers recovery assistance if requested
When The Recipe Should be Used:
- Stack traces or technical errors appear
- API response codes need explanation
- Complex validation failures occur
- Users encounter cryptic error messages
How To Use The Recipe:
# Example execution:
translated = ERROR_TRANSLATOR_RECIPE.execute({
"error_message": "TypeError: Cannot read property 'map' of undefined",
"error_context": "trying to display recipe list",
"user_expertise": "non-technical"
})
Recipe 11: COMMENT-FLOW - Communication Quality
Recipe ID: RCP-001-004-011-COMMENT-FLOW-v1.00a
Recipe Name: COMMENT-FLOW
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: AI Comment System Self-Monitor with Silent Optimization
What the Recipe Does:
Monitors AI's comment usage for consistency, ensures proper acknowledgment of H->AI directives, operates in silent mode by default to improve communication without disrupting flow.
How The Recipe Works:
- Continuously self-monitors comment usage
- Detects missed acknowledgments
- Self-corrects silently in default mode
- Can provide visible feedback if requested
- Generates usage reports on demand
- Provides comment templates when asked
When The Recipe Should be Used:
- Runs continuously in background
- When communication clarity needs improvement
- Training new CRAFT users
- Quality assurance checks
How To Use The Recipe:
# Example execution:
comment_monitor = COMMENT_WORKFLOW_RECIPE.execute({
"mode": "silent", # or "visible", "report", "templates"
"conversation_segment": "last 10 exchanges"
})
Recipe 12: EXPECT-SET - Upfront Clarity
Recipe ID: RCP-001-004-012-EXPECT-SET-v1.00a
Recipe Name: EXPECT-SET
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Proactive Expectation Management with Progress Tracker Integration
What the Recipe Does:
Communicates upfront what will happen, estimated complexity, likely duration, and known limitations before starting tasks, automatically initializes PROGRESS-TRACKER for monitoring.
How The Recipe Works:
- Analyzes task complexity
- Checks AI capabilities
- Estimates time/effort required
- Sets clear expectations
- Integrates with PROGRESS-TRACKER
- Requests confirmation before proceeding
When The Recipe Should be Used:
- Before complex multi-step tasks
- When scope is unclear
- Tasks with known limitations
- Projects requiring user actions
How To Use The Recipe:
# Example execution:
expectations = EXPECTATION_SETTER_RECIPE.execute({
"task_description": "Create complete CRAFT recipe",
"auto_init_progress": True,
"detail_level": "balanced" # or "minimal", "comprehensive"
})
Recipe 13: PROJECT-FILE-INFO-EXTRACTOR - Knowledge Preservation
Recipe ID: RCP-001-004-013-PROJECT-FILE-INFO-EXTRACTOR-v1.00a
Recipe Name: PROJECT-FILE-INFO-EXTRACTOR
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: Project File Information Extractor for Cross-Project Persistence
What the Recipe Does:
Interactive recipe that collaboratively reviews project files to identify valuable patterns, functions, and insights that should persist to the framework or cookbook files for use in future projects.
How The Recipe Works:
- Scans project files for reusable content
- Categorizes discoveries (functions, patterns, insights)
- Presents findings for collaborative evaluation
- Plans destination for each valuable item
- Resolves naming conflicts
- Generates ready-to-paste extraction report
When The Recipe Should be Used:
- At project conclusion
- After discovering useful patterns
- Before archiving projects
- During framework evolution
How To Use The Recipe:
# Example execution:
extraction = PROJECT_FILE_INFO_EXTRACTOR_RECIPE.execute({
"project_id": "CFT-PROJ-CP-022",
"review_mode": "comprehensive", # or "focused", "quick"
"include_handoffs": True
})
Recipe 14: CRAFT-BRAINSTORMER - Innovation Engine
Recipe ID: RCP-001-004-014-CRAFT-BRAINSTORMER-v1.00a
Recipe Name: CRAFT-BRAINSTORMER
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: CRAFT Framework Innovation Brainstormer
What the Recipe Does:
Structured ideation tool for generating and evaluating new CRAFT functionality ideas that align with core purposes and philosophical principles, producing scored and actionable innovation concepts.
How The Recipe Works:
- Applies brainstorming constraints (blue sky/practical/immediate)
- Generates ideas structured by CRAFT components
- Maps to core philosophy and principles
- Explains implementation approach
- Generates follow-up ideas
- Scores innovation, feasibility, and impact
When The Recipe Should be Used:
- Framework enhancement planning
- Solving persistent challenges
- Feature roadmap development
- Team brainstorming sessions
How To Use The Recipe:
# Example execution:
brainstorm = CRAFT_BRAINSTORMER_RECIPE.execute({
"focus_area": "recipes", # or "persistence", "collaboration", etc.
"num_ideas": 5,
"depth_level": "detailed",
"constraint_mode": "practical" # or "blue_sky", "immediate"
})
Recipe 15: CRAFT-OPERATIONS-MANAGER - Automatic Recipe Detection
Recipe ID: RCP-001-004-015-CRAFT-OPERATIONS-MANAGER-v1.00a
Recipe Name: CRAFT-OPERATIONS-MANAGER
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: CRAFT Operations Manager - Automatic Recipe Detector
What the Recipe Does:
Lightweight recipe that runs with every prompt to detect opportunities for using CRAFT recipes and framework components. Automatically executes appropriate recipes based on configured triggers.
How The Recipe Works:
- Analyzes each user prompt
- Checks for recipe trigger conditions
- Determines appropriate action (auto-execute, ask, warn)
- Executes recipes based on mode (production/test)
- Tracks conversation context
- Reports actions via COM comments
When The Recipe Should be Used:
- Automatically runs after CHAT-INIT
- Active throughout entire session
- No manual execution needed
- Can be disabled if requested
How To Use The Recipe:
# Example execution (usually automatic):
com_check = CRAFT_OPERATIONS_MANAGER_RECIPE.execute({
"user_prompt": "Create a blog post about CRAFT",
"conversation_context": {
"token_usage": 45,
"exchange_count": 12
},
"mode": "production" # or "test"
})
Recipe 16: CRAFT-OPERATIONS-MANAGER-TEST - Testing Mode
Recipe ID: RCP-001-004-016-CRAFT-OPERATIONS-MANAGER-TEST-v1.00a
Recipe Name: CRAFT-OPERATIONS-MANAGER-TEST
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: CRAFT Operations Manager - Test Mode
What the Recipe Does:
Test version of CRAFT-OPERATIONS-MANAGER that shows reasoning and asks before executing any recipe. Includes all functionality with additional testing logic and visibility.
How The Recipe Works:
- Runs all standard COM detection logic
- Shows reasoning for each detection
- Displays what would happen in production
- Asks permission before any execution
- Provides summary statistics
- Estimates token impact
When The Recipe Should be Used:
- Testing COM configuration
- Understanding trigger logic
- Training new users
- Debugging recipe interactions
How To Use The Recipe:
# Example execution:
com_test = CRAFT_PROMPT_TEST_RECIPE.execute({
"user_prompt": "This task will take about 5 hours",
"conversation_context": {
"token_usage": 45,
"exchange_count": 12
}
})
Recipe 17: COM-UPDATER - Configuration Tool
Recipe Name: COM-UPDATER
Note: The AI can find this recipe by referring to it as either the Recipe ID or the Recipe Name.
Title: CRAFT-OPERATIONS-MANAGER Recipe Updater - Interactive Configuration Tool
What the Recipe Does:
Systematically reviews all cookbook recipes one-by-one to update CRAFT-OPERATIONS-MANAGER trigger conditions. Uses interactive Q&A format with pauses between each recipe for user feedback.
How The Recipe Works:
- Loads all cookbook recipes
- Analyzes each recipe individually
- Proposes trigger conditions
- Asks configuration questions
- Processes user feedback
- Generates updated CRAFT-OPERATIONS-MANAGER recipe
When The Recipe Should be Used:
- Adding new recipes to COM
- Updating trigger conditions
- Reviewing COM configuration
- After cookbook updates
How To Use The Recipe:
# Example execution:
updater = COM_UPDATER_RECIPE.execute({
"starting_recipe": 1,
"include_new_recipes": True,
"output_format": "full_recipe" # or "triggers_only", "diff_view"
})
Conclusion: Your CRAFT Recipe Toolkit
The CRAFT CORE cookbook represents months of testing and refinement, distilling complex AI interaction patterns into 17 reliable, reusable recipes. Each recipe addresses specific challenges identified during the Alpha testing phases, from basic session management to sophisticated workflow tracking and error handling.
These recipes work together as an integrated system. CHAT-INIT ensures proper session setup, CRAFT-OPERATIONS-MANAGER monitors for recipe opportunities throughout the conversation, and HANDOFF_SNAPSHOT preserves context between sessions. The specialized recipes handle everything from ambiguity detection to progress tracking, creating a comprehensive framework for structured AI interaction.
As you work with CRAFT, remember that recipes are meant to be adapted and extended. While the CORE cookbook provides essential functionality, you can create your own recipes for specific domains or workflows. The standardized format ensures consistency while allowing flexibility in implementation.
Next Steps:
- Bookmark this cheatsheet for quick reference during CRAFT sessions
- Practice using the basic recipes (CHAT-INIT, HANDOFF_SNAPSHOT, AMBIGUITY-DETECT)
- Explore specialized cookbooks for domain-specific recipes
- Create your own recipes using the established patterns
- Share your experiences and improvements with the CRAFT community