skills.rest
saross
sarossCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

improve-prompt

Parse and classify seed prompts, then apply structured techniques to produce annotated improved prompts.

1|Updated Mar 7, 2026
Search Tags:#prompt-engineering#techniques#quality-assessment#llm-safety#prompt-review#satisficing#prompt-hardening
One-click install
npx skills add https://github.com/saross/personal-assistant --skill improve-prompt-saross
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: improve-prompt
Source: https://github.com/saross/personal-assistant/tree/main/skills/improve-prompt
Command: npx skills add https://github.com/saross/personal-assistant --skill improve-prompt-saross

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill closes the escape hatches that cause large language models to satisfice — producing plausible but shallow, incomplete, or unverified outputs — by systematically strengthening seed prompts with structured, verifiable instructions.

Core Features & Use Cases

  • Multi-phase hardening workflow: Parse the seed prompt, classify its task type, run a short interview to resolve ambiguities, apply selected techniques from a 16-item library, and present an annotated improved prompt.
  • Anti-satisficing techniques: Claims inventory, phase decomposition, exhaustive quantifiers, structured output, error-mode anchoring, uncertainty flagging, and more, applied in a canonical order tailored to the task.
  • Educational outputs and reuse: Produces before/after quality scores, a techniques-used table, a highlighted "Technique of the Run," and an option to save the improved prompt to a grimoire for future reuse.
  • Use case: Prepare a review or audit prompt for code, documentation, or analysis tasks to ensure exhaustive verification rather than a superficial check.

Quick Start

Paste your seed prompt and ask the assistant: Improve this prompt to be anti-satisficing, exhaustive, and verifiable.

Frequently Asked Questions about improve-prompt

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent LLM satisficing when generating analysis or extraction outputs?▼

To prevent LLM satisficing, you must harden prompts with structured phases, exhaustive quantifiers, and evidence-based checks. This Skill applies a 16-technique library to strengthen seed prompts, ensuring outputs are verifiable and exhaustive rather than superficially plausible.

What is the best way to structure a prompt for exhaustive code review and verification?▼

The best way to structure a prompt for exhaustive code review is to apply anti-satisficing techniques like phase decomposition, claims inventory, and error-mode anchoring. This Skill parses your seed prompt, classifies the task, and applies these constraints in a canonical order to force verifiable outputs.

How does prompt hardening improve output quality assessment for complex tasks?▼

Prompt hardening improves output quality assessment by applying structured constraints and a five-dimension scoring system. This Skill evaluates the improved prompt against the original, producing before/after quality scores and a techniques-used table to quantify the reduction in unverified outputs.

Can I use this prompt engineering workflow for debugging and planning tasks?▼

Yes, you can use this prompt engineering workflow for debugging, planning, generation, and transformation tasks. The Skill classifies the task type during parsing and selects appropriate anti-satisficing techniques from its library to ensure the hardened prompt enforces stronger constraints for that specific domain.

Do I need to provide a fully written prompt to start the prompt improvement process?▼

You need to provide a seed prompt to start the prompt improvement process. The Skill parses your initial input, runs a short interview to resolve ambiguities, and then selects from a 16-technique library to build a structured, annotated improved prompt with verifiable instructions.

Why does my LLM output look plausible but lack verifiable evidence during data extraction?▼

Your LLM output lacks verifiable evidence because of prompt satisficing, where the model produces shallow results without exhaustive checks. This Skill closes those escape hatches by applying uncertainty flagging, structured output requirements, and claims inventory to force evidence-based extraction.

Related Skills

Explore More Software Engineering Skills→
zjunlpzjunlp
1.1k

scienceworld-planting-coordinator

Move a seed into a soil-and-water container and resolve ambiguous targets.

Official
Basic
AkallabetAkallabet

prompt-engineering-patterns

Apply few-shot learning, chain-of-thought, and structured output patterns to LLM prompts.

Community
Advanced
AlejandroFiginiAlejandroFigini

full-output-enforcement

Enforce complete code and content generation without truncation or placeholders.

Community
Advanced
ejolietejoliet

senior-prompt-engineer

Design prompts, evaluate LLM outputs, and build agentic workflows.

Community
Advanced