lesson

Save user corrections as confidence-weighted behavioral rules that resurface before similar work.

3|Updated Oct 25, 2025
One-click install
npx skills add https://github.com/Prismaibrowser/prismspace-web --skill lesson-prismaibrowser
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: lesson
Source: https://github.com/Prismaibrowser/prismspace-web/tree/main/.agents/skills/lesson
Command: npx skills add https://github.com/Prismaibrowser/prismspace-web --skill lesson-prismaibrowser

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Repeated mistakes across sessions waste time and erode trust. This Skill captures corrections and hard-won rules as durable lessons with confidence scores, so the same guidance resurfaces automatically before similar future work instead of being forgotten. ## Core Features & Use Cases - Confidence-Weighted Lessons: Each saved rule carries a confidence score that strengthens when the same content is saved again and decays when unused, separating repeated corrections from one-off noise. - Context-Triggered Recall: Lessons are scoped to a trigger situation and optionally to a project, so memory_lesson_recall surfaces the right rule before work of the same type begins. - Use Case: After a CI pipeline hangs because bare vitest entered watch mode, save the rule "Run vitest with --run in CI contexts" so every future session touching CI scripts recalls it before running tests. ## Quick Start Ask the assistant to learn a rule from your correction, for example: "Learn this: always run vitest with --run in CI or the pipeline hangs."

Frequently Asked Questions about lesson

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make an AI remember a correction permanently?▼

Save the correction as a lesson using memory_lesson_save with the rule as content, the trigger situation as context, and a confidence score. The lesson persists across sessions and is recalled before similar future work.

What is the difference between a memory and a lesson?▼

Memories store facts and decisions, while lessons store behavior as imperative rules with confidence scores. Lessons strengthen when the same content is saved repeatedly and decay when unused, making them suited for recurring corrections.

How does lesson confidence scoring work?▼

Confidence starts at 0.7 for a direct user correction or 0.5 for a self-observed pattern. Saving identical content again strengthens the existing lesson instead of creating a duplicate, and unused lessons decay over time.

Can lessons be scoped to a specific project or repository?▼

Yes, set the project parameter when saving a lesson to scope it to a specific repo. Omit the project field for universal rules that should apply across all projects and sessions.

How do I delete a lesson saved by mistake?▼

Use memory_lesson_delete to remove a lesson saved in error. Before saving, the distilled rule is echoed back for confirmation so you can veto a bad distillation before it persists.