Some_claude_skills automatic-stateful-prompt-improver

Automatically intercepts and optimizes prompts using the prompt-learning MCP server. Learns from performance over time via embedding-indexed history. Uses APE, OPRO, DSPy patterns. Activate

install

source · Clone the upstream repo

git clone https://github.com/curiositech/some_claude_skills

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/curiositech/some_claude_skills "$T" && mkdir -p ~/.claude/skills && cp -r "$T/.claude/skills/automatic-stateful-prompt-improver" ~/.claude/skills/curiositech-some-claude-skills-automatic-stateful-prompt-improver && rm -rf "$T"

manifest: .claude/skills/automatic-stateful-prompt-improver/SKILL.md

source content

Automatic Stateful Prompt Improver

MANDATORY AUTOMATIC BEHAVIOR

When this skill is active, I MUST follow these rules:

Auto-Optimization Triggers

I AUTOMATICALLY call

mcp__prompt-learning__optimize_prompt

BEFORE responding when:

Complex task (multi-step, requires reasoning)
Technical output (code, analysis, structured data)
Reusable content (system prompts, templates, instructions)
Explicit request ("improve", "better", "optimize")
Ambiguous requirements (underspecified, multiple interpretations)
Precision-critical (code, legal, medical, financial)

Auto-Optimization Process

1. INTERCEPT the user's request
2. CALL: mcp__prompt-learning__optimize_prompt
   - prompt: [user's original request]
   - domain: [inferred domain]
   - max_iterations: [3-20 based on complexity]
3. RECEIVE: optimized prompt + improvement details
4. INFORM user briefly: "I've refined your request for [reason]"
5. PROCEED with the OPTIMIZED version

Do NOT Optimize

Simple questions ("what is X?")
Direct commands ("run npm install")
Conversational responses ("hello", "thanks")
File operations without reasoning
Already-optimized prompts

Learning Loop (Post-Response)

After completing ANY significant task:

1. ASSESS: Did the response achieve the goal?
2. CALL: mcp__prompt-learning__record_feedback
   - prompt_id: [from optimization response]
   - success: [true/false]
   - quality_score: [0.0-1.0]
3. This enables future retrievals to learn from outcomes

Quick Reference

Iteration Decision

Factor	Low (3-5)	Medium (5-10)	High (10-20)
Complexity	Simple	Multi-step	Agent/pipeline
Ambiguity	Clear	Some	Underspecified
Domain	Known	Moderate	Novel
Stakes	Low	Moderate	Critical

Convergence (When to Stop)

Improvement < 1% for 3 iterations
User satisfied
Token budget exhausted
20 iterations reached
Validation score > 0.95

Performance Expectations

Scenario	Improvement	Iterations
Simple task	10-20%	3-5
Complex reasoning	20-40%	10-15
Agent/pipeline	30-50%	15-20
With history	+10-15% bonus	Varies

Anti-Patterns

Over-Optimization

What it looks like	Why it's wrong
Prompt becomes overly complex with many constraints	Causes brittleness, model confusion, token waste
Instead: Apply Occam's Razor - simplest sufficient prompt wins

Template Obsession

What it looks like	Why it's wrong
Focusing on templates rather than task understanding	Templates don't generalize; understanding does
Instead: Focus on WHAT the task requires, not HOW to format it

Iteration Without Measurement

What it looks like	Why it's wrong
Multiple rewrites without tracking improvements	Can't know if changes help without metrics
Instead: Always define success criteria before optimizing

Ignoring Model Capabilities

What it looks like	Why it's wrong
Assumes model can't do things it can	Over-scaffolding wastes tokens
Instead: Test capabilities before heavy prompting

Reference Files

Load for detailed implementations:

File	Contents
`references/optimization-techniques.md`	APE, OPRO, CoT, instruction rewriting, constraint engineering
`references/learning-architecture.md`	Warm start, embedding retrieval, MCP setup, drift detection
`references/iteration-strategy.md`	Decision matrices, complexity scoring, convergence algorithms

Goal: Simplest prompt that achieves the outcome reliably. Optimize for clarity, specificity, and measurable improvement.