pull-learnings

Pull Learnings Command

Imports organization patterns from a shared repository into local TOON format.

Prerequisites

• CLAUDE_ORG_ID environment variable must be set
• Must run git pull first to get latest shared patterns
• Local .learnings/ directory must exist

Process

1. Read shared patterns: Load $CLOSEDLOOP_WORKDIR/.closedloop-ai/learnings/org-patterns.json
2. Convert JSON → TOON: For LLM consumption
3. Regenerate local IDs: Maintain unique IDs within local file
4. Skip echo patterns: Exclude patterns that originated from this project
5. Merge into local: Update $CLOSEDLOOP_WORKDIR/.learnings/org-patterns.toon

Echo Prevention

To avoid circular pattern propagation:

1. Read sources.json to identify pattern origins
2. Skip patterns where source_project matches current project
3. This prevents: Project A → Shared → Project A (echo)

JSON to TOON Conversion

# JSON
{
  "id": "P-001",
  "scope": "org",
  "category": "pattern",
  "trigger": "auth_flow",
  "confidence": "high",
  "seen_count": 5,
  "success_rate": 0.85,
  "flags": "",
  "applies_to": ["impl-subagent"],
  "summary": "Check token expiry"
}

# TOON
P-001|org|pattern|auth_flow|high|5|0.85||impl-subagent,"Check token expiry"

Merge Strategy

When merging into local patterns:

1. New patterns: Append to file with new local ID
2. Existing patterns (same trigger): Update metadata, keep higher seen_count
3. Conflicting patterns: Prefer pattern with higher success_rate

Local ID Generation

Local IDs are regenerated during pull to:

• Avoid collisions with existing local patterns
• Maintain consistent ID sequence in local file
• Original shared ID preserved in metadata comment

Usage

# First, pull latest from remote
git pull origin main

# Then pull learnings (via ClosedLoop orchestrator)
# Requires CLAUDE_ORG_ID to be set
export CLAUDE_ORG_ID="my-organization"

Output

Updates $CLOSEDLOOP_WORKDIR/.learnings/org-patterns.toon with:

• New patterns from shared repository
• Updated metadata for existing patterns
• Comments indicating pattern sources