# Datalad Log

> >

- **Type:** Skill
- **Install:** `agentstack add skill-neuroanalytics-data-science-harness-datalad-log`
- **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/datalad-cli/skills/datalad-log

## Install

```sh
agentstack add skill-neuroanalytics-data-science-harness-datalad-log
```

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

## About

# Skill: datalad-log

Browse the recorded history of a DataLad dataset: list run commits, inspect what a
specific run recorded, and trace which command produced a given output file.

## Steps

### Browsing run history

1. **Verify DataLad context** — check for `.datalad/` in the current directory:
   ```bash
   ls .datalad/ 2>/dev/null
   ```
   If not in a dataset, stop and inform the user.

2. **List recorded runs** — show the run history with:
   ```bash
   datalad log --oneline
   ```
   Or to show only `datalad run` commits (excludes saves, checkpoints, manual commits):
   ```bash
   git log --oneline --grep="datalad run"
   ```
   Present the output as a numbered list showing commit SHA and message.

3. **Inspect a specific run** — when the user wants detail on a particular commit:
   ```bash
   datalad rerun --report 
   ```
   This shows the run record (command, inputs, outputs, message) without re-executing.
   Report:
   - The original command
   - Declared inputs (`-i`)
   - Declared outputs (`-o`)
   - Commit message
   - Timestamp

### Tracing output provenance

When the user asks "what command produced ``":

1. **Find commits that touched the file**:
   ```bash
   git log --oneline -- 
   ```

2. **Check each commit for a run record**:
   ```bash
   datalad rerun --report 
   ```
   If the commit is a `datalad run` commit, this shows the full run metadata.
   If it is a plain save or checkpoint, report that instead.

3. **Report the full provenance chain** — if the file was produced by a run, show:
   - Which commit recorded it
   - The command that produced it
   - The declared inputs to that command

### Identifying checkpoint commits

The auto-checkpoint hook creates commits with messages like:
```
[datalad] checkpoint 2026-03-12T14:05:22Z: code/analysis.py outputs/result.csv
```
These are **not** run records — they are auto-saves. They will appear in `git log` and
`datalad log` output. To show only `datalad run` commits (excludes checkpoints and saves):
```bash
git log --oneline --grep="\[datalad run\]"
```
Note: using `--invert-grep` with multiple `--grep` flags uses OR logic and will not
correctly exclude checkpoints — use the single `--grep="\[datalad run\]"` form instead.

## Constraints

- Never re-execute a run when the user only wants to inspect it — always use
  `datalad rerun --report` for inspection, not `datalad rerun` alone.
- Always distinguish between `datalad run` commits (provenance records) and plain save
  commits or checkpoint commits in your report.
- If `datalad log` is not available (older DataLad), fall back to
  `git log --oneline --grep="datalad run"` and explain the fallback.
- For file-level differences between two commits, direct the user to `/datalad-diff`.

## 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-datalad-log
- 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%.