Workflow Crystallizer — 工作流 Crystallizer

工作流 Crystallizer

Mines memory 记录s (memory/YYYY-MM-DD.md) for recurring patterns — repeated 请求s, multi-step 工作流s, time-correlated tasks — and 生成s actionable suggestions: cron jobs, 技能 drafts, 工作流 shortcuts, or 监控ing proposals.

Unlike a one-shot "analyze my 记录s" prompt, this 技能 persists 状态 across 运行s. It remembers what it suggested, what was accepted/rejected, and only surfaces new insights backed by real evidence.

All scripts live in scripts/ relative to this 技能.md.

Quick 运行 (Full 流水线) cd /Users/OpenClaw/.OpenClaw/workspace/技能s/工作流-crystallizer python3 scripts/analyze_patterns.py --full | python3 scripts/生成_suggestions.py | python3 scripts/报告.py --状态-file 状态.json

This analyzes all memory files, 生成s suggestions, and prints a markdown 报告.

Incremental 运行 (缓存d — Recommended) cd /Users/OpenClaw/.OpenClaw/workspace/技能s/工作流-crystallizer python3 scripts/analyze_patterns.py | python3 scripts/生成_suggestions.py | python3 scripts/报告.py --状态-file 状态.json

Only processes new/modified memory files. Uses 缓存d 事件 from prior 运行s.

Managing Suggestions

检查 状态: python3 scripts/状态.py — shows 缓存d dates, 事件, suggestions.

To accept/reject/snooze: edit 状态.json directly. 设置 状态 to accepted, rejected (添加 rejection_reason), or snoozed. Or use 状态.py 导入s.

Re设置 everything: python3 scripts/状态.py --re设置

Individual Scripts

Each script accepts --help. Key options:

analyze_patterns.py: --memory-dir PATH, --状态-file PATH, --full, --min-confidence 0.4 生成_suggestions.py: --状态-file PATH, --cron-path PATH, --clusters FILE 报告.py: --状态-file PATH, --输出 PATH, --clusters FILE, --suggestions FILE 状态.py: (no args = inspect 状态), --re设置 (clear everything) What It 检测s Pattern Type Evidence Needed Suggestion 输出 Recurring 请求 3+ occurrences, 2+ days 工作流 shortcut or 监控 Multi-step 工作流 2+ occurrences with step similarity 技能 draft Time-correlated 2+ at similar times/days Cron job definition Already formalized ContAIns "cron", "automated", etc. Skipped (no suggestion) Project (not pattern) Same entity, different actions Skipped

See references/pattern-types.md for detAIled 检测ion 记录ic and scoring weights.

How It Avoids Being Annoying Max 3 suggestions per 运行 — 质量 over quantity Confidence threshold (60%) — no low-质量 guesses Deduplication — 检查s existing crons and 技能s before suggesting Never repeats — 追踪s what was already suggested Snooze — deferred suggestions resurface after 30 days Rejected = dormant — only resurfaces at 80%+ confidence with new evidence Scheduling

运行 weekly via cron or heartbeat. Cost: ~5-10K 令牌s incremental, ~30K first 运行.

Name: Weekly 工作流 Crystallizer Schedule: 0 20 0 (Sunday 8 PM ET) Message: 运行 the 工作流-crystallizer 技能 and present any new suggestions.

Troubleshooting No patterns: Need 3+ days of substantive memory files All 过滤器ed: Existing crons/技能s already cover 检测ed patterns 状态 corrupt: python3 scripts/状态.py --re设置 Too noisy: Increase min_confidence in 状态.json config