Does it automatically verify existing project documentation?

Yes, it verifies claims in CLAUDE.md against the current codebase, checking for valid file paths, commands, and function references to prevent documentation rot.

Can it help with team-wide architectural standards?

It flags high-confidence, generalizable patterns for promotion to organizational records (like a Handbook), facilitating the transition from project-specific fixes to company-wide standards.

How does it improve future coding sessions?

By updating CLAUDE.md and project documentation with durable facts and architectural decisions, it ensures future AI sessions have accurate context on patterns and project-specific gotchas.

What is a decision trace in this context?

A decision trace is a structured record of a choice made during development, capturing the alternatives considered, the reasoning, confidence levels, and the final outcome.

When does the Compound Learnings skill activate?

It typically triggers during the ship phase after a PR is created or significant work is completed, but it ignores trivial changes like typos or version bumps.

Compound Learnings

Name: Compound Learnings
Author: Brite-Nites

byBrite-Nites

0•

生产力与工作流

Captures and persists durable architectural knowledge and decision traces after completing development tasks to improve future session context.

Compound Learnings serves as the long-term memory for your engineering projects, ensuring that every development session builds upon the last like compound interest. It activates during the shipping phase to analyze code diffs, extract high-confidence decision traces, and update critical project documentation such as CLAUDE.md and ADRs. By distinguishing between fleeting session noise and durable architectural patterns, it prevents knowledge rot and provides AI models with the precise, verified context needed to maintain codebase consistency and avoid repeating past mistakes.

主要功能

01Automated extraction of structured decision traces from development workflows

02Continuous verification and auto-fixing of stale claims in CLAUDE.md

03Generation of session summaries for auto-memory and future context anchoring

040 GitHub stars

05Cross-referencing with Company and Architecture Decision Records (CDR/ADR)

06Intelligent promotion flagging for reusable architectural patterns and precedents

使用场景

01Maintaining an evergreen CLAUDE.md with accurate file paths, build commands, and conventions

02Building a searchable index of architectural precedents and trade-offs for team-wide knowledge sharing

03Ensuring AI agents remain context-aware of previous decisions and project-specific gotchas

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add brite-nites/brite-claude-plugins compound-learnings

For use in Claude.ai and ChatGPT

主要功能

01Automated extraction of structured decision traces from development workflows

02Continuous verification and auto-fixing of stale claims in CLAUDE.md

03Generation of session summaries for auto-memory and future context anchoring

040 GitHub stars

05Cross-referencing with Company and Architecture Decision Records (CDR/ADR)

06Intelligent promotion flagging for reusable architectural patterns and precedents

使用场景

01Maintaining an evergreen CLAUDE.md with accurate file paths, build commands, and conventions

02Building a searchable index of architectural precedents and trade-offs for team-wide knowledge sharing

03Ensuring AI agents remain context-aware of previous decisions and project-specific gotchas