simplicio-learn

Extract durable lessons from run trajectories and write them to persistent memory files.

9|Updated May 8, 2026
One-click install
npx skills add https://github.com/simpletibr/simplicio-loop-marketing --skill simplicio-learn-simpletibr
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: simplicio-learn
Source: https://github.com/simpletibr/simplicio-loop-marketing/tree/main/.claude/skills/simplicio-learn
Command: npx skills add https://github.com/simpletibr/simplicio-loop-marketing --skill simplicio-learn-simpletibr

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? After a coding run or session ends, the lessons it produced are lost, so the next run repeats the same mistakes and re-derives the same solutions. This Skill mines a finished run's trajectory for high-signal lessons and persists them to AGENTS.md or machine-readable memory so future runs start smarter. ## Core Features & Use Cases - Correction Mining: Detects failed-then-succeeded command pairs, classifies the error type, and records wrong-pattern to right-pattern mappings. - Precedent & Bug Pattern Storage: Stores solved-problem fingerprints and structured root-cause entries in .orchestrator/patterns.jsonl so recurring issues are reused, not re-diagnosed. - Incremental, Deduped Memory: Processes only new trajectory segments via an index file, dedups semantically, and caps memory sections at ~12 bullets with eviction. - Use Case: After simplicio-tasks finishes its self-audit, run this Skill to write the top corrections and stable workspace facts into AGENTS.md so the next session pre-empts known failures. ## Quick Start Ask the agent to run a retrospective on this session and write any durable lessons to the project memory.

Frequently Asked Questions about simplicio-learn

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

FAQPage Schema
How do I make an AI coding agent remember lessons between sessions?

Run a retrospective step that mines the session trajectory for corrections, solved precedents, and stable facts, then writes them to a persistent file like AGENTS.md. This Skill automates that with incremental indexing and deduplication so memory stays lean.

What kinds of lessons should be stored in agent memory?

Store three durable categories: command corrections with error classification, solved-problem precedents with fingerprints, and stable workspace facts or user preferences. One-off state, transcripts, and secrets are explicitly excluded.

How does the skill avoid duplicating lessons already in memory?

It loads an incremental index of previously processed trajectory segments and semantically dedups new candidates against stored bullets. Near-duplicates bump an occurrence count instead of adding new entries, and sections are capped at about 12 bullets.

When should I run a retrospective on an agent run?

Run it after a task workflow finishes its self-audit, at session end via a stop hook, or whenever the user asks to remember something. If nothing durable surfaced, the correct output is to write nothing.

Can stored lessons override the agent's safety rules?

No. Transcript and item content is treated as untrusted, and a lesson cannot encode an instruction that overrides safety gates. Memory is also bounded, deduped, and evictable, and wrong lessons are deleted rather than kept.