Agent skill

error_recognition_and_correction

Stars 0
Forks 0

Install this agent skill to your Project

npx add-skill https://github.com/colinalexander/buffet/tree/main/skills/error_recognition_and_correction

SKILL.md

SKILL: Error Recognition and Correction

Summary

Error Recognition and Correction is the skill of identifying when a judgment, assumption, or decision is wrong, and responding by updating beliefs and actions promptly and proportionately. It institutionalizes intellectual honesty and treats learning as a core risk-control mechanism.

This skill ensures the system adapts without ego or defensiveness.


Judgment Role

This skill functions as a self-correction mechanism.

It governs how the system responds to disconfirming evidence, preventing small mistakes from compounding into large losses and ensuring that learning occurs symmetrically for successes and failures.


Judgment Checkpoints

Checkpoint 1: Disconfirming Evidence Scan

Purpose:
Ensure that evidence contradicting prior judgments is actively sought and evaluated.

Key Questions:

  • What evidence would prove this judgment wrong?
  • Has such evidence emerged?
  • Are signals being dismissed or rationalized?

Required Evidence / Inputs:

  • Updated performance data
  • New qualitative or structural information

Expected Outputs:

  • Judgment Record documenting evidence assessment

Failure Modes Guarded Against:

  • Confirmation bias
  • Narrative entrenchment

Escalation Criteria:

  • Disconfirming evidence is repeatedly ignored

Checkpoint 2: Error Attribution Analysis

Purpose:
Distinguish between bad outcomes due to error and those due to randomness.

Key Questions:

  • Was the original reasoning flawed?
  • Did assumptions fail, or did outcomes vary within expectations?
  • Would the decision be repeated under similar conditions?

Required Evidence / Inputs:

  • Original judgment records
  • Outcome vs expectation comparison

Expected Outputs:

  • Judgment Record identifying error sources

Failure Modes Guarded Against:

  • Outcome bias
  • Blaming randomness selectively

Escalation Criteria:

  • Inability to articulate why an outcome occurred

Checkpoint 3: Corrective Action Discipline

Purpose:
Ensure that recognized errors lead to appropriate action.

Key Questions:

  • What adjustment is warranted?
  • Is the response proportional to the error?
  • Does correction improve future decisions?

Required Evidence / Inputs:

  • Proposed corrective actions
  • Impact assessment

Expected Outputs:

  • Judgment Record recommending adjustment, escalation, or exit

Failure Modes Guarded Against:

  • Overcorrection
  • Inertia after recognition

Escalation Criteria:

  • Errors acknowledged without corrective action

Authority Boundaries

  • Humans: Full authority; judgment-intensive.
  • Models: Advisory only; may surface anomalies and inconsistencies.
  • Prohibited: Models may not enforce corrections autonomously.

This skill may override prior decisions without stigma.


Time Horizon

  • Primary horizon: Continuous
  • Secondary horizon: Post-decision review cycles

Error recognition should occur promptly but calmly.


Interaction With Other Skills

Upstream (inputs from):

  • all decision-making skills

Downstream (feeds into):

  • capital_allocation_judgment
  • permanent_capital_loss_avoidance
  • narrative_resistance

Auditability & Records

  • Errors and corrections recorded explicitly
  • Rationales preserved for learning
  • Periodic review of repeated errors
  • Records retained permanently

Common Misuse Patterns

  • Admitting errors verbally but not behaviorally
  • Overreacting to short-term noise
  • Treating learning as retrospective only

Versioning

  • Skill version: 1.0
  • Last reviewed: 2026-01-09
  • Change notes: Initial canonical definition

Notes

The cost of being wrong is not the mistake itself, but the refusal to recognize and correct it.

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results