self-correct-reasoning

Original：🇺🇸 English
Translated

Analyze and correct previous responses when questioned or when contradictions are detected. Use this skill when the user challenges your reasoning, points out inconsistencies, or asks 'what makes you think that?' to help you review your logic, identify errors in your previous statements, and provide accurate corrections. Useful for maintaining consistency, admitting mistakes, and rebuilding trust through transparent self-evaluation.

2installs
Sourcedalehurley/phpbot
Added on

NPX Install

npx skill4agent add dalehurley/phpbot self-correct-reasoning

Tags

Translated version includes tags in frontmatter
self-correctionreasoning-improvementai-skillerror-detectiontrust-building

SKILL.md Content

View Translation Comparison →

Skill: self-correct-reasoning

When to Use

Use this skill when:
  • The user questions your reasoning with phrases like 'what makes you think that?', 'why do you say that?', or 'how did you get that?'
  • You detect contradictions between your current statement and previously retrieved data
  • You realize you made an unsupported claim or logical error
  • The user challenges the accuracy or consistency of your response
  • You need to review and correct a previous answer

Input Parameters

ParameterRequiredDescriptionExample
user_challenge
YesThe user's question or statement challenging your previous responsewhat makes you think that?
previous_response
YesYour earlier statement or claim that is being questionedno raincoat needed
relevant_context
NoPreviously retrieved data, facts, or information relevant to the claimSydney weather forecast showing light drizzle and patchy rain

Procedure

  1. Identify the specific claim or response being questioned
  2. Review the conversation history and any data you previously retrieved or stated
  3. Check for contradictions between your claim and the actual data or facts presented
  4. Identify the logical error, unsupported assumption, or inconsistency
  5. Acknowledge the error explicitly and explain what went wrong
  6. Provide the correct answer based on the actual data or sound reasoning
  7. Apologize if appropriate and thank the user for catching the mistake

Example

Example requests that trigger this skill:
what makes you think that?