Claude-skill-registry-data meta-prompt-framework
install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry-data
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry-data "$T" && mkdir -p ~/.claude/skills && cp -r "$T/data/meta-prompt-framework" ~/.claude/skills/majiayu000-claude-skill-registry-data-meta-prompt-framework && rm -rf "$T"
manifest:
data/meta-prompt-framework/SKILL.mdsource content
Meta-Prompt Framework Skill
This skill implements a systematic approach to prompt engineering and task execution optimization.
When This Skill Activates
- Designing prompts for complex tasks
- Implementing multi-phase workflows
- Optimizing LLM task execution
- Creating role-specific instructions
- Building iterative refinement loops
Three-Phase Task Flow
Phase 1: Initialization
objectives: - Analyze task context - Clarify overall goal - Generate initial outline - Prepare inputs for processing deliverables: - Task interpretation - Requirements list - Preliminary approach instruction: | Prepare inputs for deeper processing and draft an initial outline of the solution structure.
Phase 2: Refinement
objectives: - Refine initial outputs - Identify inconsistencies - Explore alternatives - Ensure modularity deliverables: - Quality-reviewed output - Alternative approaches - Edge cases identified instruction: | Refine the output to meet quality standards, considering modularity and adaptability.
Phase 3: Completion
objectives: - Finalize against criteria - Review constraints - Verify structure - Generate final output deliverables: - Complete output - Compliance check - Quality metrics instruction: | Generate the final output, ensuring clarity, accuracy, and adherence to the task flow.
Success Criteria
| Criterion | Check |
|---|---|
| Accuracy | Factual correctness verified |
| Consistency | Replicable across executions |
| Clarity | Simple and understandable |
| Efficiency | Completed in reasonable time |
Refinement Strategies
Evaluate-Apply-Feedback Loop
- Evaluate outputs for accuracy/consistency
- Apply corrections and improvements
- Generate feedback for context update
- Iterate until criteria met
Context Evolution
- Update context files with new insights
- Document improvements for future reference
- Feed learnings back into prompts
Role Separation
Task Executor
- Generates outputs following task flow
- Implements three-phase model
- Focuses on deliverable creation
Task Reviewer
- Reviews for clarity and accuracy
- Checks meta-directive adherence
- Provides improvement suggestions
Templates
Phase Template
## Phase [N]: [Name] **Objective:** [What to accomplish] **Inputs:** - [Input 1] - [Input 2] **Process:** 1. [Step 1] 2. [Step 2] **Outputs:** - [Deliverable 1] - [Deliverable 2] **Success Criteria:** - [ ] [Criterion 1] - [ ] [Criterion 2]
Meta-Directive Template
meta_directives: [directive_name]: priority: [high|medium|low|critical] action: "[specific action to take]" constraint: "[limitation to observe]"
Best Practices
- Phase Gates: Complete each phase before proceeding
- Explicit Criteria: Define success metrics upfront
- Role Clarity: Separate execution from review
- Iteration Limits: Set maximum refinement cycles
- Context Preservation: Maintain state across phases