# Run Comparison

> >

- **Type:** Skill
- **Install:** `agentstack add skill-neuroanalytics-data-science-harness-run-comparison`
- **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/run-comparison

## Install

```sh
agentstack add skill-neuroanalytics-data-science-harness-run-comparison
```

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

## About

# Skill: run-comparison

Run the user's analysis script as a provenanced, containerized computation so inputs, command,
container image, and outputs are all recorded (Actionability + Tracking) in a disposable,
rebuilt-from-spec environment (Portability + Ephemerality). You never run the tool yourself —
you delegate to the **datalad doer**.

## When to use
- A comparison branch exists (`analyze/propose-comparison`) and the analysis script is written.
- Do NOT use for one-off exploratory commands that produce no output files — those need no run
  record. Do NOT use to create the branch (that is `propose-comparison`).

## Steps
1. **Confirm context** — verify you are on the comparison's branch (`cmp/`) and the script
   exists in `code/`. If not, ask the user which branch/script, or route to `propose-comparison`.
2. **Gather run parameters**:
   - `command` — how to invoke the script (e.g. `python code/stats.py`)
   - `-i` inputs — the data files/globs the script reads (e.g. `participants.tsv`, `sub-*/…`)
   - `-o` outputs — where results land (prefer `derivatives/cmp-/…`)
   - `-m` message — a meaningful description of the run
   - container — the project container recipe from `containers/` (build/register on first use)
3. **Ensure the container image exists (containers doer)** — if no `.sif` has been built yet for the
   project recipe, delegate to the **containers doer**:
   > "build a `.sif` from `containers/` into `containers/.sif`."
   It returns the image path and the `datalad containers-add …` command to register it. Skip if the
   image is already built and registered.
4. **Delegate execution to the datalad doer**:
   > "container-run `` on branch `cmp/` using container `` (register it with the
   > `containers-add` command from step 3 if not yet registered), inputs ``, outputs ``,
   > message ``."
   Wait for its structured result (commit sha, recorded outputs, pass/fail).
5. **Handle the result**:
   - **ok** — note the commit and output paths.
   - **failed** — relay the doer's error + suggested fix; nothing was committed. Stop.
6. **Log it** — append to `project.yaml`:
   `{ ts, op: run-comparison, stage: analyze, note: "container-run ; commit ;
   outputs ", branch: cmp/ }`.
7. **Report** — commit, outputs, and that the run is replayable via the doer (`datalad rerun`).

## Constraints
- Always run through the datalad doer's `container-run` path so the environment is captured — do
  not fall back to a bare `datalad run` or a raw shell command for an analysis with outputs.
- Require a meaningful `-m` message; never a placeholder.
- Do not write or "fix" the analysis script's science — if it errors on its own logic, surface it
  to the user; the harness owns provenance, not the model.
- Keep `project.yaml` append-only; log both successful and (as a note) failed runs if useful.

## 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-run-comparison
- 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%.