Skip to content
Back to skills

MxSave

ASecurity

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.

  • 30 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 25, 2026
ai-agentsgobashgit

Works with

  • mcp

Security analysis

A100/100

Pro scans all 16 files and shows the line behind each finding

Scanned September 25, 2026

npx -y skills add MicrotronX/mxLore --skill mxSave --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of MxSave?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for MxSave
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/microtronx-mxsave/badge)](https://www.skillsdirectory.com/skills/microtronx-mxsave)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: mxSave
description: 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.
user-invocable: true
effort: medium
allowed-tools: Read, Write, Edit, Grep, Glob, Bash, Task, AskUserQuestion
argument-hint: "[optional-notes] [--loop] [--delta-check]"
---

# /mxSave — Persist Project State (AI-Steno: !=forbidden →=use ⚡=critical ?=ask)

Thin dispatcher. Each step: read its `references/` file BEFORE executing it (verbatim procedure there). ⚡ Delegating a step to a subagent → put the ABSOLUTE path (skill base dir + `references/<file>`) of that step's file and of every file it names (e.g. `lesson-template.json`) into the spawn prompt; a subagent never sees this skill's base dir.

## Execution Mode ⚡
Phased: Main 1,1b,2 → parallel A (3 bg-subagent sonnet + 4a Main) → Main 5 sync → Main 4b + 5b → B: 6 fire-and-forget. !single fan-out. Context/Init/phases/degraded path → `references/init-and-execution-mode.md`.

## Steps
- **Init** (slug, load deferred tools, mx_ping→`mcp_available`, state-file safety) → `references/init-and-execution-mode.md`
- **1** settings.local.json clean, **1b** artifact sweep (report-only) → `references/step1-settings-artifacts.md`
- **2** CLAUDE.md + status.md (byte gates, `note#PENDING`, zombie check) → `references/step2-claude-status.md`
- **3** MCP docs (WF cleanup, stale sweep, archive, FR/BR closure in Main, lessons, findings) → `references/step3-mcp-docs.md`
- **4a/4b** orchestrate state sync → `references/step4-orchestrate-sync.md`
- **5** session note (Main, sync) → `references/step5-session-note.md` + `references/body-validation.md`
- **5b** resume handoff → `references/step5b-handoff.md`
- **6** peer notify + Final Block → `references/step6-final-block.md`
- Modes `--delta-check` (alias `--clear-cycle`), `--loop` → `references/modes-delta-check-loop.md`
- Rules + Completion output → `references/rules-completion.md`

## Invariants (every run) ⚡
- ⚡ `last_save_deltas = state_deltas` set ONLY in 4a (single writer), BEFORE reset; one deferred state.json write in 4b.
- ⚡ No done-marker before the proven side effect (handoff only after note doc_id + 4b write landed).
- ⚡ Handoff ignore gate: `git check-ignore` exit 1 → no write, delete existing.
- ⚡ Status-entry discipline: exactly ONE `**Status:**` entry + ONE History pointer; replace, never accumulate. Pointers exactly once, verify greps ASCII-only.
- ⚡ status.md: remove a block only if it carries `#NNNN`.
- ⚡ now_utc from `date -u`, never the chat clock.
- ⚡ Note sections `Quickstart after /clear`, `Blocked on me`, `Tooling gotchas + verify` never omitted; `status='active'`.
- ⚡ NO internal reasoning in notes/summary; facts only, confirmed-implemented only.
- ⚡ Batch-dismiss of findings forbidden; mxSave writes no verdicts.
- ⚡ Prompts via AskUserQuestion; never auto-archive FR/BR/TODO — ONE exception: items carrying the complete `fixed-pending-verify` marker past its deadline (Step 3, rules in `~/.claude/skills/_shared/backlog-hygiene.md`).
- ⚡ `--delta-check` is not a save.

Files in this skill

  • SKILL.md3.3 KB
  • references/body-validation.md4 KB
  • references/init-and-execution-mode.md2.7 KB
  • references/lesson-template.json705 B
  • references/modes-delta-check-loop.md3.1 KB
  • references/pruning.md8.1 KB
  • references/rules-completion.md1 KB
  • references/state-write-semantics.md1.6 KB
  • references/step-state-sync.md3.7 KB
  • references/step1-settings-artifacts.md1.5 KB
  • references/step2-claude-status.md11.7 KB
  • references/step3-mcp-docs.md8.7 KB
  • references/step4-orchestrate-sync.md7.7 KB
  • references/step5-session-note.md3.4 KB
  • references/step5b-handoff.md2 KB
  • references/step6-final-block.md3.4 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.