| name | thread-checkpoint |
| description | Capture a compact, resume-ready checkpoint of the current Codex session so work survives a long thread, a context reset, or a handoff. Use when the conversation is getting long, before switching tasks, when the user says "checkpoint", "save where we are", "summarize so we can continue later", or "hand this off". Writes a durable status note — not chat fluff — that the next session can act on cold.
|
Thread Checkpoint
Long sessions drift and context gets summarized away. A checkpoint preserves the decisions and state, not the narration, so work can resume without re-deriving everything.
What to capture
Write a tight note with these sections — short, specific, and grounded in what actually happened:
- Objective — the goal in one or two sentences. What "done" means.
- State — what is true right now: branch, what's deployed/committed/local-only, services running, key file paths touched.
- Done — the concrete things finished, each verifiable (a commit, a passing test, a deployed URL).
- Open / next — the remaining steps as an ordered list of commands or actions, not vague intentions. Put the single next action first.
- Decisions — choices already settled and why, so they aren't relitigated (e.g. "using X over Y because …").
- Blockers — anything waiting on the user or an external system, named precisely.
Where to write it
- Prefer a file so it survives compaction:
CHECKPOINT.md at the repo root, or update the project's status doc if one exists.
- Convert relative time to absolute dates.
- Keep it under ~30 lines. If it's longer, you're narrating instead of checkpointing.
Resuming from one
When handed a checkpoint, read it fully, then verify the State against the real repo/runtime before acting — treat it as context, not proof the world still matches it.