learn
Capture a learning pattern from the current session with confidence scoring.
Skill by eroslifestyle
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/eroslifestyle-learn-b3e340
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/learn
Tools it pre-approves 4
From allowed-tools in SKILL.md: the agent may use these without asking you each time.
Read,Write,Edit,Glob
What's inside
- Skill name
learn- Description
- Capture a learning pattern from the current session with confidence scoring. Use when the user says /learn or when a reusable pattern is discovered.
- License
- MIT
- Allowed tools
- Read,, Write,, Edit,, Glob
- Instructions
- 126 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Learn Skill - Continuous Learning System
## Purpose
Capture reusable patterns discovered during sessions as "instincts" with confidence scoring.
Instincts start weak (0.3) and grow stronger with repeated confirmation (max 0.9).
## Trigger
- User types `/learn` followed by a pattern description
- User types `/learn` with no args (auto-detect from current session context)
## Algorithm
1. **Parse input**: Extract the pattern description and optional tags from user input
2. **Read instincts file**: Load `~/.claude/learnings/instincts.json`
- If file is empty or missing, initialize with default structure
3. **Check for duplicates**: Compare new pattern against existing instincts
- Fuzzy match: if 60%+ of significant words (4+ chars) overlap, treat as existing
- If match found: increment confidence by 0.2 (cap at 0.9), add new evidence entry, update last_confirmed
- If no match: create new instinct with confidence 0.3
4. **Generate instinct object** (for new patterns):
```json
{
"id": "inst_NNN",
"pattern": "<pattern description>",
"confidence": 0.3,
"evidence": ["session_YYYY-MM-DD: <context>"],
"created": "YYYY-MM-DD",
"last_confirmed": "YYYY-MM-DD",
"tags": ["<auto-detected-or-user-provided>"],
"source": "user|auto",
"evolved_to": null
}
```
5. **Save**: Write updated instincts.json, increment metadata.total_learned
6. **Display**: Show the saved/updated instinct to the user in a clear format
## ID Generation
- Format: `inst_NNN` where NNN is zero-padded 3-digit sequential number
- Read existing instincts, find max ID number, increment by 1
- First instinct: `inst_001`
## Tag Auto-Detection
If user does not provide tags, infer from pattern text using these keyword groups:
| Keywords in pattern | Tag |
|---|---|
| parallel, concurrent, async, simultaneous | `parallelism` |
| glob, grep, search, find | `search` |
| performance, fast, speed, optimize | `performance` |
| test, assert, verify, check | `testing` |
| error, fix, bug, debug | `debugging` |
| git, commit, branch, merge | `git` |
| file, read, write, edit | `file-ops` |
| security, auth, token, secret | `security` |
| cache, memory, resource | `resources` |
Assign up to 3 tags. If no keywords match primary categories, try secondary:
| Keywords | Tag |
|---|---|
| checkpoint, session, memory, context | `session-management` |
| architecture, design, structure | `architecture` |
| workflow, process, pipeline | `workflow` |
If still no match, use `["general"]` (last resort).
## Fuzzy Match Logic
```
significant_words(text) = [word.lower() for word in text.split() if len(word) >= 4]
# Jaccard similarity (intersection/union) is symmetric and handles
# different-length patterns better than precision-style overlap
jaccard(a, b) = len(set(significant_words(a)) & set(significant_words(b))) / len(set(significant_words(a)) | set(significant_words(b)))
is_duplicate = jaccard >= 0.6
```
## Confidence Scale
| Observation | Confidence | Meaning |
|---|---|---|
| 1st (creation) | 0.3 | New pattern, first discovery |
| 2nd (confirmed once) | 0.5 | Same pattern seen again (+0.2) |
| 3rd (confirmed twice) | 0.7 | Evolution candidate (+0.2, 3+ confirms) |
| 4th+ (confirmed 3+) | 0.9 | Verified pattern (+0.2, capped at 0.9) |
Formula: `new_confidence = min(current_confidence + 0.2, 0.9)`
Contradiction: `new_confidence = max(current_confidence - 0.2, 0.0)` -> archive at 0.0
## Output Format
```
INSTINCT CAPTURED
ID: inst_NNN
Pattern: <description>
Confidence: 0.X [NEW | +0.2 CONFIRMED]
Tags: tag1, tag2
Evidence: N entries
Tip: Use /evolve when you have 3+ instincts at 0.7+ to create reusable skills.
```
## Corruption Detection and Recovery
**Detection criteria:** instincts.json is corrupted if:
- Invalid JSON syntax
- Missing "version" field
- Missing "instincts" array
- File size > 10MB
**Recovery:**
1. Back up as `instincts.json.YYYYMMDD_HHMMSS.bak` (timestamped, not overwrite)
2. Keep max 3 backups (delete oldest)
3. Reinitialize with fresh schema
4. Log: "Recovered from corrupted instincts.json"From eroslifestyle/Claude-Orchestrator-Plugin, MIT.
More from eroslifestyle/Claude-Orchestrator-Plugin
- 0.432score
evolve
eroslifestyle
Aggregate high-confidence instincts into reusable skills.
SkillDFreshness: D★ 1Claude Code +9 - 0.432score
debugging
eroslifestyle
When to use this skill for systematic debugging including root cause analysis, error tracing, performance profiling, and memory leak detection
SkillDFreshness: D★ 1Claude Code +9 - 0.431score
code-review
eroslifestyle
When to use this skill for automated code quality analysis and review
SkillDFreshness: D★ 1Claude Code +9 - 0.431score
git-workflow
eroslifestyle
When to use this skill for advanced Git operations and workflow management
SkillDFreshness: D★ 1Claude Code +9 - 0.429score
plan
eroslifestyle
Decompose a feature request into a detailed implementation plan with task breakdown, dependencies, and risk assessment.
SkillDFreshness: D★ 1Claude Code - 0.429score
prompt-engineering-patterns
eroslifestyle
Master advanced prompt engineering techniques to maximize LLM performance, reliability, and controllability.
SkillDFreshness: D★ 1Claude Code +9 - 0.429score
python-patterns
eroslifestyle
Python best practices, idioms, and patterns.
SkillDFreshness: D★ 1Claude Code +9 - 0.429score
go-patterns
eroslifestyle
Go best practices, idioms, and patterns.
SkillDFreshness: D★ 1Claude Code +9 - 0.426score
cleanup
eroslifestyle
Run session cleanup - delete temp files, NUL files, orphan artifacts
SkillDFreshness: D★ 1Claude Code +9 - 0.426score
checkpoint
eroslifestyle
Save a named checkpoint of current session progress.
SkillDFreshness: D★ 1Claude Code +9 - 0.426score
metrics
eroslifestyle
Display detailed session metrics and agent performance.
SkillDFreshness: D★ 1Claude Code +9 - 0.426score
multi-plan
eroslifestyle
Create a multi-agent execution plan for complex tasks.
SkillDFreshness: D★ 1Claude Code
Signals
- Stars
- 1
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Mar 3, 2026 (7 months ago)
- Last release
- Mar 3, 2026 (7 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
Same skill elsewhere
Copies of this SKILL.md published by others; search shows only one of them.
More from eroslifestyle
- 0.432score
evolve
eroslifestyle
Aggregate high-confidence instincts into reusable skills.
SkillDFreshness: D★ 1Claude Code +9 - 0.432score
debugging
eroslifestyle
When to use this skill for systematic debugging including root cause analysis, error tracing, performance profiling, and memory leak detection
SkillDFreshness: D★ 1Claude Code +9 - 0.431score
git-workflow
eroslifestyle
When to use this skill for advanced Git operations and workflow management
SkillDFreshness: D★ 1Claude Code +9 - 0.431score
code-review
eroslifestyle
When to use this skill for automated code quality analysis and review
SkillDFreshness: D★ 1Claude Code +9 - 0.429score
python-patterns
eroslifestyle
Python best practices, idioms, and patterns.
SkillDFreshness: D★ 1Claude Code +9 - 0.429score
go-patterns
eroslifestyle
Go best practices, idioms, and patterns.
SkillDFreshness: D★ 1Claude Code +9
Cite this page
learn skill. Agent Mart. https://agentmart.dev/skills/eroslifestyle-learn-b3e340 (last updated 2026-10-24)