continuous-learninglisted

# Continuous Learning Scan archived session logs to surface recurring manual workflows, then offer to convert high-frequency patterns into new SKILL.md stubs so repeated work becomes permanent automation. ## Usage ``` /continuous-learning # Scan all archived sessions for patterns /continuous-learning --min 2 # Lower threshold (default: 3 repetitions) /continuous-learning --generate # Auto-generate stubs without prompting ``` ## Behavior ### Phase 1: Discover sessions ```bash # Find all archived session files (not the active .current-session pointer) find .claude/sessions -name "*.md" -not -name ".current-session" 2>/dev/null | sort ``` If fewer than 3 session files exist, report "Not enough session history yet (need 3+)" and exit. ### Phase 2: Extract commands and workflows Scan each session file for lines that match invocation patterns: ```bash # Commands invoked (slash commands, skill names) grep -h -E "^/[a-z]|^\*\*Command\*\*:|Ran:|Invoked:" .claude/sessions/**/*.md 2>/dev/null \ | sed 's|.*: ||' | sort | uniq -c | sort -rn | head -40 ``` Also extract multi-step sequences: look for clusters of 3+ consecutive commands that appear together in at least `--min` sessions. ### Phase 3: Build pattern registry Load or create `.claude/cache/continuous-learning/patterns.json`: ```json { "last_scan": "2026-06-04T00:00:00Z", "patterns": [ { "commands": ["/security-scan", "/review", "/test"], "frequency": 5, "sessions": ["s