aiblueprint-cli 1.4.22 → 1.4.24

package/claude-code-config/skills/meta-prompt-creator/SKILL.md

@@ -0,0 +1,285 @@
+---
+name: prompt-creator
+description: Expert prompt engineering for creating effective prompts for Claude, GPT, and other LLMs. Use when writing system prompts, user prompts, few-shot examples, or optimizing existing prompts for better performance.
+---
+
+<objective>
+Create highly effective prompts using proven techniques from Anthropic and OpenAI research. This skill covers all major prompting methodologies: clarity, structure, examples, reasoning, and advanced patterns.
+
+Every prompt created should be clear, specific, and optimized for the target model.
+</objective>
+
+<quick_start>
+<workflow>
+
+1. **Clarify purpose**: What should the prompt accomplish?
+2. **Identify model**: Claude, GPT, or other (techniques vary slightly)
+3. **Select techniques**: Choose from core techniques based on task complexity
+4. **Structure content**: Use XML tags (Claude) or markdown (GPT) for organization
+5. **Add examples**: Include few-shot examples for format-sensitive outputs
+6. **Define success**: Add clear success criteria
+7. **Test and iterate**: Refine based on outputs
+   </workflow>
+
+<core_structure>
+Every effective prompt has:
+
+```xml
+<context>
+Background information the model needs
+</context>
+
+<task>
+Clear, specific instruction of what to do
+</task>
+
+<requirements>
+- Specific constraints
+- Output format
+- Edge cases to handle
+</requirements>
+
+<examples>
+Input/output pairs demonstrating expected behavior
+</examples>
+
+<success_criteria>
+How to know the task was completed correctly
+</success_criteria>
+```
+
+</core_structure>
+</quick_start>
+
+<core_techniques>
+<technique name="be_clear_and_direct">
+**Priority**: Always apply first
+
+- State exactly what you want
+- Avoid ambiguous language ("try to", "maybe", "generally")
+- Use "Always..." or "Never..." instead of "Should probably..."
+- Provide specific output format requirements
+
+See: [references/clarity-principles.md](references/clarity-principles.md)
+</technique>
+
+<technique name="use_xml_tags">
+**When**: Claude prompts, complex structure needed
+
+Claude was trained with XML tags. Use them for:
+
+- Separating sections: `<context>`, `<task>`, `<output>`
+- Wrapping data: `<document>`, `<schema>`, `<example>`
+- Defining boundaries: Clear start/end of sections
+
+See: [references/xml-structure.md](references/xml-structure.md)
+</technique>
+
+<technique name="few_shot_examples">
+**When**: Output format matters, pattern recognition easier than rules
+
+Provide 2-4 input/output pairs:
+
+```xml
+<examples>
+<example number="1">
+<input>User clicked signup button</input>
+<output>track('signup_initiated', { source: 'homepage' })</output>
+</example>
+</examples>
+```
+
+See: [references/few-shot-patterns.md](references/few-shot-patterns.md)
+</technique>
+
+<technique name="chain_of_thought">
+**When**: Complex reasoning, math, multi-step analysis
+
+Add explicit reasoning instructions:
+
+- "Think step by step before answering"
+- "First analyze X, then consider Y, finally conclude Z"
+- Use `<thinking>` tags for Claude's extended thinking
+
+See: [references/reasoning-techniques.md](references/reasoning-techniques.md)
+</technique>
+
+<technique name="system_prompts">
+**When**: Setting persistent behavior, role, constraints
+
+System prompts set the foundation:
+
+- Define Claude's role and expertise
+- Set constraints and boundaries
+- Establish output format expectations
+
+See: [references/system-prompt-patterns.md](references/system-prompt-patterns.md)
+</technique>
+
+<technique name="prefilling">
+**When**: Enforcing specific output format (Claude-specific)
+
+Start Claude's response to guide format:
+
+```
+Assistant: {"result":
+```
+
+Forces JSON output without preamble.
+</technique>
+
+<technique name="context_management">
+**When**: Long-running tasks, multi-session work, large context usage
+
+For Claude 4.5 with context awareness:
+
+- Inform about automatic context compaction
+- Add state tracking (JSON, progress.txt, git)
+- Use test-first patterns for complex implementations
+- Enable autonomous task completion across context windows
+
+See: [references/context-management.md](references/context-management.md)
+</technique>
+</core_techniques>
+
+<prompt_creation_workflow>
+<step_0>
+**Gather requirements** using AskUserQuestion:
+
+1. What is the prompt's purpose?
+   - Generate content
+   - Analyze/extract information
+   - Transform data
+   - Make decisions
+   - Other
+
+2. What model will use this prompt?
+   - Claude (use XML tags)
+   - GPT (use markdown structure)
+   - Other/multiple
+
+3. What complexity level?
+   - Simple (single task, clear output)
+   - Medium (multiple steps, some nuance)
+   - Complex (reasoning, edge cases, validation)
+
+4. Output format requirements?
+   - Free text
+   - JSON/structured data
+   - Code
+   - Specific template
+     </step_0>
+
+<step_1>
+**Draft the prompt** using this template:
+
+```xml
+<context>
+[Background the model needs to understand the task]
+</context>
+
+<objective>
+[Clear statement of what to accomplish]
+</objective>
+
+<instructions>
+[Step-by-step process, numbered if sequential]
+</instructions>
+
+<constraints>
+[Rules, limitations, things to avoid]
+</constraints>
+
+<output_format>
+[Exact structure of expected output]
+</output_format>
+
+<examples>
+[2-4 input/output pairs if format matters]
+</examples>
+
+<success_criteria>
+[How to verify the task was done correctly]
+</success_criteria>
+```
+
+</step_1>
+
+<step_2>
+**Apply relevant techniques** based on complexity:
+
+- **Simple**: Clear instructions + output format
+- **Medium**: Add examples + constraints
+- **Complex**: Add reasoning steps + edge cases + validation
+  </step_2>
+
+<step_3>
+**Review checklist**:
+
+- [ ] Is the task clearly stated?
+- [ ] Are ambiguous words removed?
+- [ ] Is output format specified?
+- [ ] Are edge cases addressed?
+- [ ] Would a person with no context understand it?
+      </step_3>
+      </prompt_creation_workflow>
+
+<anti_patterns>
+<pitfall name="vague_instructions">
+❌ "Help with the data"
+✅ "Extract email addresses from the CSV, remove duplicates, output as JSON array"
+</pitfall>
+
+<pitfall name="negative_prompting">
+❌ "Don't use technical jargon"
+✅ "Write in plain language suitable for a non-technical audience"
+</pitfall>
+
+<pitfall name="no_examples">
+❌ Describing format in words only
+✅ Showing 2-3 concrete input/output examples
+</pitfall>
+
+<pitfall name="missing_edge_cases">
+❌ "Process the file"
+✅ "Process the file. If empty, return []. If malformed, return error with line number."
+</pitfall>
+
+See: [references/anti-patterns.md](references/anti-patterns.md)
+</anti_patterns>
+
+<reference_guides>
+**Core principles:**
+
+- [references/clarity-principles.md](references/clarity-principles.md) - Being clear and direct
+- [references/xml-structure.md](references/xml-structure.md) - Using XML tags effectively
+
+**Techniques:**
+
+- [references/few-shot-patterns.md](references/few-shot-patterns.md) - Example-based prompting
+- [references/reasoning-techniques.md](references/reasoning-techniques.md) - Chain of thought, step-by-step
+- [references/system-prompt-patterns.md](references/system-prompt-patterns.md) - System prompt templates
+- [references/context-management.md](references/context-management.md) - Context windows, long-horizon reasoning, state tracking
+
+**Best practices by vendor:**
+
+- [references/anthropic-best-practices.md](references/anthropic-best-practices.md) - Claude-specific techniques
+- [references/openai-best-practices.md](references/openai-best-practices.md) - GPT-specific techniques
+
+**Quality:**
+
+- [references/anti-patterns.md](references/anti-patterns.md) - Common mistakes to avoid
+- [references/prompt-templates.md](references/prompt-templates.md) - Ready-to-use templates
+  </reference_guides>
+
+<success_criteria>
+A well-crafted prompt has:
+
+- Clear, unambiguous objective
+- Specific output format with example
+- Relevant context provided
+- Edge cases addressed
+- No vague language (try, maybe, generally)
+- Appropriate technique selection for task complexity
+- Success criteria defined
+  </success_criteria>

package/claude-code-config/skills/meta-prompt-creator/references/anthropic-best-practices.md

@@ -0,0 +1,126 @@
+<overview>
+Claude-specific prompting techniques from Anthropic's official documentation.
+</overview>
+
+<techniques>
+<technique name="xml_tags">
+Claude was trained with XML tags. Use them for structure:
+- `<context>` - Background information
+- `<instructions>` - Task description
+- `<document>` - Content to process
+- `<example>` - Input/output pairs
+</technique>
+
+<technique name="be_explicit">
+State exactly what you want:
+- Use "Always..." or "Never..." instead of "try to"
+- Specify format, length, style explicitly
+- Add context about WHY (helps Claude make better decisions)
+</technique>
+
+<technique name="extended_thinking">
+For complex reasoning:
+- Add "Think step by step"
+- Guide: "First analyze X, then consider Y, finally conclude Z"
+</technique>
+
+<technique name="prefilling">
+Start Claude's response to enforce format:
+- Begin with `{"result":` for JSON
+- Prevents preamble text
+</technique>
+
+<technique name="positive_framing">
+State what TO DO, not what NOT to do:
+- BAD: "Don't use jargon"
+- GOOD: "Write in plain language for non-technical readers"
+</technique>
+
+<technique name="prompt_chaining">
+Break complex tasks into steps:
+- Output of one prompt becomes input for next
+- Each step has clear, focused objective
+</technique>
+</techniques>
+
+<claude_4_specific>
+<extended_thinking>
+- Leverage extended thinking for complex tasks
+- Use trigger phrases: "Thoroughly analyze...", "Consider multiple approaches..."
+- Extended thinking tokens count toward context but are auto-stripped from subsequent turns
+</extended_thinking>
+
+<parallel_tools>
+- Direct parallel tool calls explicitly
+- "Invoke all independent operations simultaneously"
+- Improves performance for file reads, searches, API calls
+</parallel_tools>
+
+<investigation_first>
+- Instruct to investigate/read files BEFORE answering
+- "Read existing implementation before proposing changes"
+- Prevents hallucination and ensures context-aware responses
+</investigation_first>
+
+<context_awareness>
+**Claude Sonnet 4.5 and Haiku 4.5** have built-in context awareness:
+- Track remaining token budget throughout conversation
+- Receive periodic updates on capacity
+- Manage long-running tasks more effectively
+
+For systems with automatic context compaction (like Claude Code):
+```xml
+<context>
+Your context window will be automatically compacted as it approaches
+its limit, allowing you to continue working indefinitely from where
+you left off.
+</context>
+```
+
+This prevents Claude from stopping work prematurely.
+</context_awareness>
+
+<long_horizon_tasks>
+For tasks spanning multiple context windows:
+
+**State tracking:**
+- Use structured formats (JSON) for task status
+- Use unstructured notes (progress.txt) for context
+- Leverage git for work logs and checkpoints
+
+**Instructions pattern:**
+```xml
+<instructions>
+Work systematically through all tasks. Track progress in:
+- progress.json (structured status)
+- progress.txt (session notes)
+- git commits (implementation history)
+
+If context refreshes, review these files to resume.
+</instructions>
+```
+
+**Test-first pattern:**
+```xml
+<requirements>
+Create all tests in tests.json first. It is unacceptable to
+remove or edit tests - they are the source of truth.
+</requirements>
+```
+
+See: [context-management.md](context-management.md) for comprehensive patterns.
+</long_horizon_tasks>
+
+<context_window_management>
+**Size limits:**
+- Standard: 200K tokens
+- Enterprise: 500K tokens
+- Sonnet 4/4.5 (beta): 1M tokens
+
+**Best practices:**
+- Plan for context accumulation
+- Use token counting API for estimates
+- Consider premium pricing (>200K = 2x input, 1.5x output)
+- Images count toward budget
+</context_window_management>
+</claude_4_specific>

package/claude-code-config/skills/meta-prompt-creator/references/anti-patterns.md

@@ -0,0 +1,57 @@
+<overview>
+Common prompt mistakes and how to avoid them.
+</overview>
+
+<anti_patterns>
+<pitfall name="vague_instructions">
+BAD: "Help with the data"
+GOOD: "Extract emails from CSV, remove duplicates, output as JSON array"
+</pitfall>
+
+<pitfall name="negative_prompting">
+BAD: "Don't use technical jargon"
+GOOD: "Write in plain language for non-technical readers"
+
+Negative instructions can backfire - state what to do instead.
+</pitfall>
+
+<pitfall name="no_examples">
+BAD: Describing format in words only
+GOOD: Showing 2-3 concrete input/output examples
+
+Examples communicate nuances words cannot.
+</pitfall>
+
+<pitfall name="missing_edge_cases">
+BAD: "Process the file"
+GOOD: "Process the file. If empty, return []. If malformed, return error with line number."
+</pitfall>
+
+<pitfall name="ambiguous_language">
+BAD: "Try to keep it short", "Maybe add examples"
+GOOD: "Maximum 3 paragraphs", "Include 2 examples"
+</pitfall>
+
+<pitfall name="no_success_criteria">
+BAD: "Analyze the data"
+GOOD: "Analyze the data. Success: identify top 3 trends with supporting metrics."
+</pitfall>
+
+<pitfall name="too_many_options">
+BAD: "You can use library A, B, C, D, or E..."
+GOOD: "Use library A. For edge case X, use B instead."
+
+Provide one default with escape hatch.
+</pitfall>
+
+<pitfall name="inconsistent_terminology">
+BAD: Mixing "API endpoint", "URL", "route", "path"
+GOOD: Pick one term and use consistently throughout
+</pitfall>
+</anti_patterns>
+
+<testing>
+Ask: "Could I hand these instructions to someone with no context and expect correct results?"
+
+If unclear to a human, it's unclear to the model.
+</testing>

package/claude-code-config/skills/meta-prompt-creator/references/clarity-principles.md

@@ -0,0 +1,54 @@
+<overview>
+Clarity reduces errors and improves output quality.
+</overview>
+
+<golden_rule>
+Show your prompt to someone with minimal context. If they're confused, the model will be too.
+</golden_rule>
+
+<guidelines>
+<guideline name="specificity">
+BAD: "Help with the report"
+GOOD: "Generate markdown report with: Executive Summary, Key Findings, Recommendations"
+</guideline>
+
+<guideline name="sequential_steps">
+Provide numbered instructions:
+1. Extract data
+2. Transform format
+3. Validate
+4. Save output
+</guideline>
+
+<guideline name="avoid_ambiguity">
+Replace ambiguous phrases:
+- "Try to..." → "Always..."
+- "Should probably..." → "Must..."
+- "Generally..." → "Always... except when..."
+</guideline>
+
+<guideline name="define_edge_cases">
+Anticipate edge cases:
+- What if no results?
+- What if duplicates?
+- What if invalid format?
+</guideline>
+
+<guideline name="output_format">
+Specify format with example:
+```json
+{"name": "string", "email": "string"}
+```
+</guideline>
+
+<guideline name="success_criteria">
+Define success:
+- All rows parsed
+- No validation errors
+- Output file created
+</guideline>
+</guidelines>
+
+<show_dont_tell>
+When format matters, show an example rather than describing it.
+</show_dont_tell>