← ClaudeAtlas

mxsavelisted

Use when the user says "save state", "/mxSave", "session end", "before /compact", "wrap up", or otherwise wants to persist the current mx-project state (clean settings, update CLAUDE.md + docs/status.md, create session notes in MCP-DB, sync orchestrate-state deltas, emit clear-cycle tip). Loop-capable. Fires at natural session-end boundaries.
MicrotronX/mxLore · ★ 25 · AI & Automation · score 78
Install: claude install-skill MicrotronX/mxLore
# /mxSave — Persist Project State (AI-Steno: !=forbidden →=use ⚡=critical ?=ask) > **Context:** Hybrid — MCP work→background subagent, `.claude/` files→Main (subagents lack write perm). Long bodies (>300 words) via background subagent; mx_detail default 600 tokens. Save agent. Persists project state for seamless session continuation. **Hybrid:** CLAUDE.md+status.md=local. Session notes=MCP-DB. ## Execution Mode ⚡ **Phased sequential-with-parallel** (do NOT collapse into a single fan-out; race + data-dependency hazards): 1. **Main:** Init → Steps 1, 2 (settings + CLAUDE.md/status.md + zombie check). 2. **Parallel phase A:** Step 3 (background subagent, MCP-only) + Step 4a (Main, in-memory mutations). Pass `mcp_available` to Step 3 explicitly; Step 4a sends `expected_updated_at` and skips WFs already archived by Step 3. 3. **Main (synchronous):** Step 5 — `mx_create_doc(session_note)` issued from Main; subagent may build the body string but Main issues the call and captures the doc_id (skill runtime has no await-subagent primitive — running Step 5 in background would regress Bug#3229). 4. **Main:** Step 4b — single deferred Write applying ALL 4a + 4b mutations (incl. `last_save_summary` + `last_save_session_note_doc_id` from Step 5's return). 5. **Parallel phase B (fire-and-forget):** Step 6 Peer Notify — no join, errors logged not aborted. Degraded path: Step 5 MCP call fails → Step 4b writes `last_save_summary` (local) + `last_save_session_note_doc_id=null`. ## Init 1.