Skip to content

Instantly share code, notes, and snippets.

@kkurian
Created February 2, 2026 21:34
Show Gist options
  • Select an option

  • Save kkurian/bf1f8e2e417bb7eeee7a630d6f1547c5 to your computer and use it in GitHub Desktop.

Select an option

Save kkurian/bf1f8e2e417bb7eeee7a630d6f1547c5 to your computer and use it in GitHub Desktop.
High-Confidence Methodology - Claude Code Skill for rigorous iterative investigation
name description
high-confidence
This skill should be used when the user indicates they want high confidence in findings or research results. Implements an iterative investigation loop where every significant finding must reach high confidence through verification before being accepted.

High-Confidence Methodology

Apply rigorous iterative investigation to reach high confidence on every significant finding.

Arguments

  • $ARGUMENTS - (Optional) The research question, topic, or task to investigate with high confidence. If not provided, apply the methodology to whatever is currently being worked on in the conversation.

Activation

When this skill is invoked, immediately use the Skill tool to start a ralph-loop with the high-confidence completion promise:

  • Skill name: ralph-loop:ralph-loop
  • Arguments: $ARGUMENTS --completion-promise 'All significant findings have reached high confidence'

If no arguments are provided, use the current task context as the prompt for the ralph-loop arguments.

Core Loop

For each significant finding, apply this loop:

  1. Assess confidence: After forming a conclusion, ask "How confident am I in this?"
  2. If not high confidence: Ask "What is the next question I need answered to increase my confidence?"
  3. Investigate: Take action to answer that question.
  4. Repeat until high confidence is reached.

If high confidence cannot be reached after reasonable investigation, explicitly note what remains uncertain and why.

Completion

When all significant findings have reached high confidence, output:

<promise>All significant findings have reached high confidence</promise>

Only output this when it is TRUE. Do not output it prematurely to exit the loop.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment