# Checkpoint

> >

- **Type:** Skill
- **Install:** `agentstack add skill-neuroanalytics-data-science-harness-checkpoint`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [neuroanalytics](https://agentstack.voostack.com/s/neuroanalytics)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [neuroanalytics](https://github.com/neuroanalytics)
- **Source:** https://github.com/neuroanalytics/data-science-harness/tree/main/plugins/analyze/skills/checkpoint

## Install

```sh
agentstack add skill-neuroanalytics-data-science-harness-checkpoint
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Skill: checkpoint

Take a clean, described snapshot of the dataset. This keeps the provenance chain continuous (a
dirty working tree produces misleading run records downstream). You delegate the save to the
**datalad doer**.

> v1 note: checkpoint is **on-demand** — invoked by the user or suggested by the coordinator. An
> automatic end-of-session `datalad save` (a `Stop`/session-end hook) is intentionally deferred:
> how often vs. how long to trigger it is a latency/tuning question that likely varies per study
> (plan gap B5), best decided after real use.

## When to use
- The user is pausing/ending a session, or wants intermediate state recorded.
- Do NOT use to record a provenanced *computation* — that is `analyze/run-comparison`
  (`container-run`), which already commits its own outputs. Checkpoint captures hand edits,
  notes, and other loose changes.

## Steps
1. **Inspect state** — delegate to the datalad doer:
   > "status: report modified/untracked files in this dataset and the current branch."
   If nothing is unsaved, tell the user there is nothing to checkpoint and stop.
2. **Compose a message** — summarize what changed since the last save into a meaningful `-m`
   (e.g. "checkpoint: draft stats.py + participant notes"). Ask the user if the change set is
   ambiguous.
3. **Save** — delegate to the datalad doer:
   > "save: `datalad save -m ''` on the current branch."
4. **Log it** — append to `project.yaml`:
   `{ ts, op: checkpoint, stage: , note: "", branch:  }`.
5. **Report** — the commit sha and current branch.

## Constraints
- Delegate status and save to the datalad doer; never call datalad directly.
- Always use a meaningful message — never "wip"/"save"/placeholder.
- Keep `project.yaml` append-only.
- Do not force-save over a comparison mid-run; if a `container-run` is in progress, let it finish
  (it commits its own outputs).

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [neuroanalytics](https://github.com/neuroanalytics)
- **Source:** [neuroanalytics/data-science-harness](https://github.com/neuroanalytics/data-science-harness)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-neuroanalytics-data-science-harness-checkpoint
- Seller: https://agentstack.voostack.com/s/neuroanalytics
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.