What is confidence decay in this skill?

Confidence decay uses a 90-day half-life formula to reduce the weight of old feedback, ensuring the swarm prioritizes recent, relevant signals over stale historical data.

How does the system score task outcomes?

It uses a weighted formula that considers success (40%), execution duration (20%), error count (20%), and retry attempts (20%) to determine if a strategy was helpful or harmful.

When does a strategy become an anti-pattern?

A strategy is automatically inverted into an anti-pattern if it has at least 3 observations and a failure rate exceeding 60%, prompting Claude to avoid that approach in the future.

What is the benefit of the Proven maturity state?

Patterns marked as 'proven' (having high helpful counts and low failure ratios) receive a 1.5x multiplier during decomposition, significantly increasing the likelihood of successful project outcomes.

Swarm Learning Systems

Name: Swarm Learning Systems
Author: primeinc

byprimeinc

Productividad y Flujo de Trabajo

Implements automated feedback loops and pattern optimization to improve AI task decomposition quality based on historical outcomes.

Acerca de

Swarm Learning Systems provides a sophisticated framework for Claude to learn from task outcomes autonomously. By tracking metrics like execution duration, error counts, and retry frequency, the system calculates weighted scores to categorize decomposition patterns as candidates, proven, or deprecated. It features a confidence decay mechanism to ensure stale patterns don't dominate and an automated anti-pattern detection system that inverts high-failure strategies into explicit 'avoid' instructions, ensuring the swarm becomes more efficient and accurate with every subtask performed.

Características Principales

0 GitHub stars
Centralized Error Accumulator for detailed retry context
Implicit feedback scoring based on duration, errors, and retries
Anti-pattern inversion for strategies with high failure rates
Automated pattern maturity lifecycle (Candidate, Established, Proven)
Confidence decay formula to prevent dominance of stale patterns

Casos de Uso

Optimizing AI task decomposition strategies for complex software projects
Debugging why specific coding patterns are being promoted or deprecated
Building self-improving AI workflows that learn from execution history

Acerca de

Características Principales

0 GitHub stars
Centralized Error Accumulator for detailed retry context
Implicit feedback scoring based on duration, errors, and retries
Anti-pattern inversion for strategies with high failure rates
Automated pattern maturity lifecycle (Candidate, Established, Proven)
Confidence decay formula to prevent dominance of stale patterns

Casos de Uso

Optimizing AI task decomposition strategies for complex software projects
Debugging why specific coding patterns are being promoted or deprecated
Building self-improving AI workflows that learn from execution history