UNPKG

claude-flow-novice

Version:

Claude Flow Novice - Advanced orchestration platform for multi-agent AI workflows with CFN Loop architecture Includes CodeSearch (hybrid SQLite + pgvector), mem0/memgraph specialists, and all CFN skills.

142 lines (111 loc) 3.91 kB
--- name: cfn-expert-update description: "Update CFN system expert agent with relevant git commits since last scan" argument-hint: "[--dry-run] [--since=commit_hash] [--force]" allowed-tools: ["Bash", "Read", "Write", "Grep", "Glob"] --- # CFN Expert Update Scan git commits since last run and update the CFN system expert agent with relevant process changes. ## Purpose Updates the `.claude/agents/custom/cfn-system-expert.md` agent with: - New CLI commands and patterns - Updated CFN Loop processes - New skills and coordination patterns - Cost optimization strategies - Adaptive context lessons ## Execution ```bash # Standard update (scans since last tracked commit) /cfn-expert-update # Dry run (show what would be updated without making changes) /cfn-expert-update --dry-run # Force scan from specific commit /cfn-expert-update --since=abc123 # Force full re-scan (ignore last commit tracking) /cfn-expert-update --force ``` ## Process ### 1. Commit Tracking - Reads last scanned commit from `.claude/state/cfn-expert-last-commit` - Scans git log from that point to HEAD - Updates tracking file after successful scan ### 2. Relevance Detection Scans commits for CFN-related patterns: - `CLAUDE.md` changes (system rules) - `.claude/commands/cfn/` additions/modifications - `.claude/skills/cfn-*` updates - Agent coordination patterns - Cost optimization updates - CFN Loop methodology changes ### 3. Knowledge Extraction For relevant commits, extracts: - New commands and usage patterns - Updated processes and workflows - Performance insights - Bug fixes and solutions - Cost optimization strategies ### 4. Agent Update Updates the expert agent with: - Enhanced command knowledge - New troubleshooting patterns - Updated best practices - Current cost comparison data - Latest adaptive context lessons ## State Management **Last Commit Tracking**: `.claude/state/cfn-expert-last-commit` ``` abc123def456789 # Commit hash of last scan 2025-01-15T10:30:00Z # Timestamp of last update ``` **Backup Strategy**: Creates backup before updates: ``` .claude/backups/cfn-expert/20250115_103000_cfn-system-expert.md ``` ## Relevance Patterns **High Priority** (always updates): - `CLAUDE.md` modifications - `/cfn-loop-*` command changes - CFN Loop methodology updates **Medium Priority** (contextual updates): - `.claude/skills/cfn-*` changes - Agent coordination improvements - Cost optimization strategies **Low Priority** (informational): - Documentation updates - Test additions - Performance improvements ## Output Examples ### Dry Run Mode ``` 🔍 Scanning commits since abc123def (5 commits found) 📋 Relevant commits: 3/5 🎯 Updates available: New command: /cfn-loop-frontend (P1: 0.95) Updated cost table (P2: 0.87) New coordination pattern (P3: 0.92) 💡 Run without --dry-run to apply updates ``` ### Update Mode ``` 🔍 Scanning commits since abc123def (5 commits found) 📋 Relevant commits: 3/5 🎯 Applying 3 updates to cfn-system-expert agent: Added /cfn-loop-frontend command knowledge Updated cost optimization strategies Enhanced coordination patterns 📁 Backup created: .claude/backups/cfn-expert/20250115_103000_* 📝 Last commit updated: def456abc789 🔄 Expert agent updated successfully ``` ## Error Handling - **No new commits**: "No new commits to scan" - **No relevant changes**: "No CFN-relevant changes found" - **Git errors**: Check repository status and permissions - **File errors**: Verify agent file exists and is writable --- ## Implementation Notes - Uses git log with commit hash filtering - Pattern matching for relevance detection - Safe updates with backup/restore capability - Idempotent operations (safe to re-run) - Timestamp tracking for audit trail