ericmjl/continuous-learning-v3

Continuous Learning v3

An agent-agnostic learning system that observes your AI coding sessions and creates reusable knowledge through atomic "instincts" — small learned behaviors with confidence scoring.

⚡ First-Time Setup (Automatic)

When this skill is first loaded, automatically perform these steps:

Step 1: Check if Plugin is Installed

# Check for the OpenCode plugin
ls ~/.config/opencode/plugins/continuous-learning.js

Step 2: If Not Installed, Install It

If the plugin file does not exist:

1. Find the skill directory (where this SKILL.md is located)
2. Copy the plugin:

   cp <skill-directory>/plugins/continuous-learning.js \
      ~/.config/opencode/plugins/continuous-learning.js
   

3. Create a marker file to track installation:

   touch ~/.config/opencode/skill/continuous-learning-v3/.installed
   

Step 3: Create Storage Directories

mkdir -p ~/.agent-learning/instincts/personal
mkdir -p ~/.agent-learning/instincts/inherited
mkdir -p ~/.agent-learning/evolved/skills
mkdir -p ~/.agent-learning/evolved/commands
mkdir -p ~/.agent-learning/evolved/agents
mkdir -p ~/.agent-learning/projects

Step 4: Tell User to Restart

After installing the plugin, tell the user:

✅ Continuous Learning plugin installed!

Please restart OpenCode to activate automatic learning.

After restart, all tool executions will be observed and patterns will be learned automatically.

To check learning status: uv run ~/.config/opencode/skill/continuous-learning-v3/scripts/instinct-cli.py status

---

When to Activate

• User asks about learning, patterns, or preferences
• User wants to see what the agent has learned
• User wants to export/import instincts
• User wants to evolve instincts into skills
• First time the skill is loaded (run setup)

The Instinct Model

An instinct is a small learned behavior:

---
id: prefer-functional-style
trigger: "when writing new functions"
confidence: 0.7
domain: "code-style"
source: "session-observation"
scope: project
project_id: "a1b2c3d4e5f6"
project_name: "my-react-app"
agent: "opencode"
---

# Prefer Functional Style

## Action
Use functional patterns over classes when appropriate.

## Evidence
- Observed 5 instances of functional pattern preference
- User corrected class-based approach on 2025-01-15

Properties:

• Atomic — one trigger, one action
• Confidence-weighted — 0.3 = tentative, 0.9 = near certain
• Domain-tagged — code-style, testing, git, debugging, workflow, etc.
• Evidence-backed — tracks what observations created it
• Scope-aware — project (default) or global
• Agent-aware — tracks which agent learned this pattern

CLI Commands

Use the CLI to manage instincts:

# Show all instincts (project + global)
uv run <skill-dir>/scripts/instinct-cli.py status

# Cluster related instincts into skills/commands
uv run <skill-dir>/scripts/instinct-cli.py evolve

# Export instincts
uv run <skill-dir>/scripts/instinct-cli.py export -o my-instincts.yaml

# Import instincts
uv run <skill-dir>/scripts/instinct-cli.py import my-instincts.yaml

# Promote project instincts to global
uv run <skill-dir>/scripts/instinct-cli.py promote

# List known projects
uv run <skill-dir>/scripts/instinct-cli.py projects

# Migrate from v2.x
uv run <skill-dir>/scripts/instinct-cli.py migrate --from ~/.claude/homunculus

File Structure

~/.agent-learning/
+-- config.yaml             # Configuration
+-- projects.json           # Registry: project hash -> name/path/remote
+-- observations.jsonl      # Global observations (fallback)
+-- instincts/
|   +-- personal/           # Global auto-learned instincts
|   +-- inherited/          # Global imported instincts
+-- evolved/
|   +-- agents/             # Global generated agents
|   +-- skills/             # Global generated skills
|   +-- commands/           # Global generated commands
+-- projects/
    +-- a1b2c3d4e5f6/       # Project hash (from git remote URL)
    |   +-- project.json    # Per-project metadata
    |   +-- observations.jsonl
    |   +-- instincts/
    |   |   +-- personal/   # Project-specific auto-learned
    |   |   +-- inherited/  # Project-specific imported
    |   +-- evolved/
    |       +-- skills/
    |       +-- commands/
    |       +-- agents/
    +-- ...                 # Other projects

Scope Decision Guide

| Pattern Type | Scope | Examples | |-------------|-------|---------| | Language/framework conventions | project | "Use React hooks", "Follow Django REST patterns" | | File structure preferences | project | "Tests in __tests__/" | | Code style | project | "Use functional style", "Prefer dataclasses" | | Security practices | global | "Validate user input", "Sanitize SQL" | | General best practices | global | "Write tests first", "Always handle errors" | | Tool workflow preferences | global | "Grep before Edit", "Read before Write" | | Git practices | global | "Conventional commits", "Small focused commits" |

Confidence Scoring

| Score | Meaning | Behavior | |-------|---------|----------| | 0.3 | Tentative | Suggested but not enforced | | 0.5 | Moderate | Applied when relevant | | 0.7 | Strong | Auto-approved for application | | 0.9 | Near-certain | Core behavior |

Environment Variables

| Variable | Default | Description | |----------|---------|-------------| | AGENT_LEARNING_HOME | ~/.agent-learning | Base directory for all data | | AGENT_PROJECT_DIR | (auto-detected) | Override project detection | | AGENT_LEARNING_DISABLED | (unset) | Set to 1 to disable observation |

Troubleshooting

Plugin Not Loading

1. Check plugin exists: ls ~/.config/opencode/plugins/continuous-learning.js
2. Restart OpenCode
3. Check for errors in OpenCode logs

No Observations Being Captured

1. Check AGENT_LEARNING_DISABLED is not set
2. Verify project detection: git remote get-url origin
3. Check observations file: cat ~/.agent-learning/projects/*/observations.jsonl

Reset Everything

rm -rf ~/.agent-learning
rm ~/.config/opencode/plugins/continuous-learning.js
rm ~/.config/opencode/skill/continuous-learning-v3/.installed
# Then reload this skill to re-run setup

Privacy

• Observations stay local on your machine
• Project-scoped instincts are isolated per project
• Only instincts (patterns) can be exported — not raw observations
• No actual code or conversation content is shared
• You control what gets exported and promoted

Attribution

This skill is a generalized, agent-agnostic fork of continuous-learning-v2 from the everything-claude-code repository.