wangchuan 4.1.0 → 4.2.0

package/skill/claude/SKILL.md

@@ -1,96 +0,0 @@
----
-name: wangchuan
-description: >
-  AI memory sync system — auto-sync configs, skills, and memories across agents and environments.
-  TRIGGER: Activate at the END of every task session to check if any configs, memories, or skills
-  should be synced. Also activate when user says "sync", "wangchuan", "push memories", "pull memories".
----
-
-# Wangchuan — AI Agent Config Sync System
-
-## When to Activate
-
-This skill MUST activate in these scenarios:
-
-1. **End of every task session** — Before concluding, check if anything valuable was learned or changed
-2. **User explicitly requests** — "sync", "wangchuan sync", "push/pull memories"
-3. **After modifying agent config files** — CLAUDE.md, settings.json, skills/, .claude.json
-4. **After learning something reusable** — New workflow patterns, user preferences, tool configurations
-
-## Auto-start Watch Daemon
-
-On skill activation, check if the watch daemon is running:
-```bash
-# Check if watch daemon PID file exists and process is alive
-if [ -f ~/.wangchuan/watch.pid ] && kill -0 $(cat ~/.wangchuan/watch.pid) 2>/dev/null; then
-  : # Watch daemon is running
-else
-  # Start watch daemon in background for continuous sync
-  nohup wangchuan watch > /dev/null 2>&1 &
-fi
-```
-This ensures continuous sync while the agent session is active.
-
-## Self-Evolution Protocol
-
-At the end of each task session, evaluate whether the conversation produced any of these:
-
-| Category | What to look for | Where to update |
-|----------|-----------------|-----------------|
-| **New skill learned** | Reusable workflow, automation pattern, or domain knowledge | Create/update skill in `~/.claude/skills/` |
-| **User preference discovered** | Coding style, tool preference, communication preference | Append to `~/.claude/CLAUDE.md` |
-| **Memory worth preserving** | Key decisions, project context, recurring patterns | Append to agent memory file |
-| **Config change made** | MCP servers added, settings changed, permissions updated | Already tracked by file sync |
-
-### Decision Criteria — Only Update When Truly Valuable
-
-Do NOT update for:
-- Trivial one-off tasks (fixing a typo, answering a quick question)
-- Information that is project-specific and not reusable
-- Temporary debugging steps
-
-DO update for:
-- Patterns the user will likely need again across projects
-- Explicit user requests ("remember this", "add this to my preferences")
-- New tool integrations or workflow automations
-- Corrections to existing memory/skills (fixing outdated info)
-
-## Commands — Just 6
-
-```bash
-# The ONE daily command — smart bidirectional sync
-wangchuan sync
-
-# Check sync state at a glance
-wangchuan status
-
-# Full detail (file list, diff, history, health)
-wangchuan status -v
-
-# Diagnose + auto-fix everything
-wangchuan doctor
-
-# Filter to specific agent
-wangchuan sync --agent claude
-
-# Multi-environment management
-wangchuan env list
-wangchuan env create work
-wangchuan env switch work
-```
-
-## Current Sync Status
-
-!`wangchuan status 2>/dev/null || echo "wangchuan not initialized"`
-
-## End-of-Session Checklist
-
-Before concluding any task:
-
-1. **Review**: Did this session produce reusable knowledge?
-2. **Evaluate**: Is it worth persisting? (Apply the decision criteria above)
-3. **Update**: If yes, update the appropriate file (memory/skill/config)
-4. **Sync**: Run `wangchuan sync` to propagate to all agents and cloud
-5. **Report**: Briefly tell the user what was synced (if anything)
-
-If nothing valuable was produced, skip silently — do not announce "nothing to sync".