State-diagram skill for Claude Code: project memory turns gotchas into explicit flow conditions
DJIRNMAN · reddit · 2026-11-02
The author added a state-diagram skill to MEX, an open-source project-memory tool for Claude Code and Codex, after finding Claude's textual flow explanations hard to picture. Unlike plain Mermaid output, MEX injects recorded project decisions and gotchas into the diagram with inspectable code references.
Demoed on MEX's own setup/recovery flow: code analysis covered normal pauses, retries, and saves, but a project note surfaced an awkward case — rebuilding the CLI while an old setup process still runs removes a module that process references, and Retry doesn't refresh its cached imports. That changes the recovery path: you need a fresh CLI process but can reuse prepared setup files; the diagram spells out the conditions (old process exited, build coherent, prepared files exist).
Diagrams can be requested in-conversation or explored in a local Hub with transitions, conditions, and evidence; the skill also supports TLA+/TLC checks on the modeled rules. Repo: github.com/mex-memory/mex
Related event: MEX adds state-diagram skill to help Claude Code visualize project logic(2 posts)→
More from coding & agent
- clusyio's Portal lets Python jobs migrate to a new machine mid-run — AIFlow_ML · 2026-11-02
- 50 agents stalled on bureaucracy instead of building in Minecraft experiment — petrusenko_max · 2026-11-02
- Tobi demos pi CLI: one prompt lets an agent download and set a system font — evilrabbit_ · 2026-11-02
- A crew of 5 Grok bots runs your job hunt on autopilot — nima_owji · 2026-11-02
- CorebrumOS Physical AI OS runs on ARM SBC with 8 built-in AI agents driving robots — chrismatthieu · 2026-11-02
- Agentic CI/CD: AI agents fix the pipeline that fails at 2 a.m. — Pavan_Belagatti · 2026-11-02