learn

Manage project learnings by reviewing, searching, pruning, and exporting patterns across sessions.

1|Updated Aug 17, 2026
One-click install
npx skills add https://github.com/NalinDalal/skillset --skill learn-nalindalal
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: learn
Source: https://github.com/NalinDalal/skillset/tree/main/skills/workflow/learn
Command: npx skills add https://github.com/NalinDalal/skillset --skill learn-nalindalal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Teams lose hard-won project knowledge between sessions, repeatedly rediscovering the same pitfalls, fixes, and conventions. This Skill persists durable learnings to a JSONL file and lets you review, search, prune, and export them. ## Core Features & Use Cases - Review and Search: Show the 20 most recent learnings grouped by type (pattern, pitfall, preference, architecture) or grep the learnings file for a query. - Prune and Export: Flag stale entries referencing deleted files or contradictory keys, then export curated learnings as formatted markdown. - Automatic Logging: After any skill completes, durable session insights are appended to .opencode/learnings.jsonl with type, key, confidence, and timestamp. - Use Case: A developer wonders "didn't we fix this before?" and runs a search to instantly retrieve the earlier pitfall and its resolution instead of re-debugging. ## Quick Start Ask the agent to show what we have learned so far on this project.

Frequently Asked Questions about learn

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

FAQPage Schema
How do I save project learnings between AI coding sessions?▼

Log each durable learning as a JSON line in .opencode/learnings.jsonl with a type, kebab-case key, one-sentence insight, confidence score, and UTC timestamp. Entries persist on disk so future sessions can read and search them.

How to search past project patterns and pitfalls?▼

Run the search command with a query, which greps .opencode/learnings.jsonl case-insensitively and returns the last 20 matching entries. Results can be grouped by type: pattern, pitfall, preference, or architecture.

Can I export project learnings as markdown documentation?▼

Yes, the export command formats all learnings as a markdown document with sections for Patterns, Pitfalls, Preferences, and Architecture, each entry showing its key, insight, and confidence score.

How do I clean up stale or contradictory learnings?▼

Use the prune command, which loads all entries and flags learnings referencing files that no longer exist or keys with conflicting insights. Each flagged entry is presented so you can choose to remove, keep, or update it.

What are the limitations of JSONL-based learning storage?▼

The JSONL file is append-only plain text with no deduplication or indexing, so pruning contradictions is a manual review step. It is also project-local, so learnings do not sync across repositories or machines.