LogoClawIndex
CasesSkillsAbout
LogoClawIndex

empirical-prompt-tuning - Iteratively improve agent-facing instructions

Iteratively improves agent-facing prompts by running unbiased executors, measuring outcomes, analyzing ambiguity, and applying fixes until improvements plateau.

Tags

Updated: 2026-10-05
prompt engineeringagent instructionsiterative evaluationsubagent evaluationinstruction testing

Capabilities

Check description-body consistencyPrepare evaluation scenariosDefine requirements checklistsDispatch blank-slate executors

Typical Inputs

Target promptEvaluation scenariosRequirements checklist

Typical Outputs

Executor self-reportsStructured issue reflectionsSuccess judgments

What this skill does

  • Check description-body consistency
  • Prepare evaluation scenarios
  • Define requirements checklists
  • Dispatch blank-slate executors
  • Evaluate results two-sidedly
  • Analyze unclear points
  • Apply minimal prompt fixes
  • Check convergence

Inputs

  • Target prompt
  • Evaluation scenarios
  • Requirements checklist
  • Failure pattern ledger
  • Task tool results

Outputs

  • Executor self-reports
  • Structured issue reflections
  • Success judgments
  • Accuracy measurements
  • Step counts
  • Duration measurements
  • Retry counts
  • Prompt diffs

Requirements

  • Task tool access
  • Blank-slate subagent dispatch
  • Task usage metadata

Source

  • Spec: SKILL.md

ClawIndex

OpenClaw Skills & Use Case Index

ClawIndex is an ecosystem-driven index of OpenClaw skills and real-world use cases.

Index

Skills·
Cases

Meta

About·
Disclaimer·
Email·
GitHub
© 2026 ClawIndex All Rights Reserved.