SMAA: Session Memory for AI Agents

Stop today, resume tomorrow — two commands, zero re-asking.

Say "stop" to close clean. Say "where we are" to continue — on any project.

Quick start · Commands · How it works · Install · FAQ · Contributing

git clone https://github.com/lab1207/smaa

---

Quick start

Give a new AI this repo and paste AI_PROMPT.md:

1. Clone SMAA next to your project.
2. Paste AI_PROMPT.md into a fresh AI session.
3. Say "where we are" to resume — "stop" when done.

The AI reads the guides, continues from Next up, and closes with everything pushed.

Commands (only 2)

You say	AI does
stop (or save everything / we stop here / end session / wrap up)	Updates PROGRESS → pushes all repos → replies Next up
where we are	Reads PLAN→PROGRESS + git → replies status → continues from Next up

Detail: COMMANDS.md. Aliases run the identical ritual — never new commands.

How it works

File	Job	Changes
PLAN.md	WHAT we build	Rarely
PROGRESS.md	WHERE we are (snapshot, Next up, log)	Every session

Trackers hold detail, PROGRESS holds the verdict. Missing SKILL-CONFIG.md = defaults. Full map: SKILL.md.

Install

One open format (Agent Skills spec) — the same SKILL.md works across 27+ agents. Per-agent commands: INSTALL.md (OpenCode, Claude Code, Cursor, Codex, project-local).

FAQ

My stack?

Any — app or website. Gate command and paths resolve per project.

My own docs layout?

Add SKILL-CONFIG.md with your paths, no renaming needed.

Contribute?

Issues and PRs welcome — one lesson per PR, generic only, no names ever. See CONTRIBUTING.md.

License

MIT