/checkpoint

Make it your own.

In Claude Code, use the slash command and add your context. In Codex, select checkpoint from the just-vibe skill picker, then send the same brief.

Version 0.11.0 also supports /jv checkpoint, /just-vibe checkpoint and /jv:checkpoint in Claude. See shortcut setup and context examples.

What the agent does

1. Save the objective, constraints, decisions, completed evidence, remaining work and next step using project checkpoint NAME with the current revision when structured storage is appropriate. Read an existing checkpoint and its revision with project resume NAME (project list shows names), and read it back the same way after saving. The helper captures repository/worktree identity per file; record each external operation ID with its confirmed or uncertain state in completed or remaining, without credentials, so resume can reconcile it before any retry.
2. If a tracked run is active, save its latest run record in the checkpoint run field so session resume keeps its stages, attempts and budget.
3. A checkpoint does not contain reversible file content. If undo support is requested before editing, create a separate task begin/capture record; never manufacture past ownership from a later snapshot.

Inputs

• current task, destination if supplied, and continuity needs.

Optional context: scope, references, constraints, successCriteria, environment, mode, budget.

Scope

Reads

Save task state in a local approved artifact; no permanent behavioral memory or external posting.

Writes

Apply: only the requested local changes and relevant isolated verification. Inspect/plan requests remain inspection/planning. External actions require their exact action and target in session authorization.

Mode

Apply; current task, destination if supplied, and continuity needs.

Prerequisites

Resolve the user brief and inspect the relevant project or supplied evidence. External capabilities are optional unless the selected action actually needs them.

Expected output

• Concise timestamped checkpoint with evidence references and one executable next action.

How the work is checked

• A later session can locate the work and distinguish done from pending; stale checks retain the revision they actually tested.

When to stop or clarify

• Do not overwrite an unrelated checkpoint. Ask about destination only when no established project location can be inferred.

Handling missing context

Infer

Resolve the named files, existing scripts, current task and earlier corrections from the conversation and repository.

Assume

Use the narrowest interpretation that completes a reversible local task; state a consequential assumption once.

Ask

Ask when competing targets or incompatible success conditions would change the result; continue independent inspection first.

Technical guidance

Evidence

Inspect worktree/index, task state, evidence, pending effects and chosen checkpoint location.

Method

Save a bounded continuation record with identities and unresolved next steps, excluding credentials and stale claims.

Pitfall

Saving notes does not capture every external effect or guarantee another host will load them.

Check

Read the saved checkpoint back and verify its artifact identities and staleness checks without overwriting unrelated memory.

Situational decisions

When a test predates intervening edits: Retain its old revision and mark current verification pending.

The coding agent follows this workflow using its available tools. Installation does not grant service access or guarantee an outcome. Read the compatibility notes.

Keep exploring