Legacy AI Session Governance

🔴  [New Standalone Version: https://github.com/Adamchanadam/agent-handoff-kit] ; Data conversion between the old and new versions is incompatible. It is recommended to use the new version of AGENT Handoff Kit.

English | 繁體中文 | 简体中文 | 日本語

:rocket: Governance Template for Cross-AI Handoff Workflows

When your Codex / Claude / Gemini quota runs out, paste the handoff block into the next AI and it picks up where you left off.

- Handoff works across different AI CLIs
- Standard workflow: PLAN -> READ -> CHANGE -> QC -> PERSIST
- Keeps governance from drifting, instead of just adding more rules
- A harness engineering component focused on session continuity

How a session works · Install · Upgrade · Quick Operations
🆕 First time here? The Interactive Introduction walks you through what this template does in about 5 minutes — recommended before reading the rest of this README.

:bookmark_tabs: Why this exists

With multiple AI tools, handoff is usually what breaks first — not output quality.

Common failure pattern:
- Context resets every time you switch tools
- Fixes pile on top of fixes, rules get messier
- README, handoff docs, and logs stop matching

This template requires:
1. One re-entry path every session
2. One workflow for every task
3. One persistent record before closing

:bookmark_tabs: Built-in safeguards

It also catches a few common AI mistakes:

:small_blue_diamond: How SESSION_LOG.md stays manageable

dev/SESSION_LOG.md is read at every session start. In an active project this file can grow to thousands of lines — loading months of history that has no relevance to today's work.

The template handles this with explicit closeout checks (not memory-only rules):

- At closeout, AI checks whether SESSION_LOG.md exceeds 400 lines or contains entries older than…