# Propose Comparison

> >

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

## Install

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

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

## About

# Skill: propose-comparison

Create a comparison: a small, branch-scoped analysis unit realized as a **named DataLad branch**
plus a project-log entry. Comparisons span a rigor spectrum: an **exploratory** quick query (default)
that lives only as a branch/run, or a **confirmatory** one whose spec is frozen and registered
*before* execution via `govern/preregister`. The *record* here is already the spec-centric research
object of STAMPED §3.12.1.

## When to use
- The user wants to run any analysis/plot/comparison against the dataset.
- Do NOT use to initialize a project (`project/new-project`) or to execute an already-proposed
  comparison (`analyze/run-comparison`).

## Steps
1. **Capture the comparison** — get from the user (short, one-liner each):
   - `what` — the comparison in a sentence ("group difference in outcome Y")
   - `why` — the question it answers *(optional but encouraged)*
   - expected `inputs` and `outputs` (best guess; refined at run time)
   - rigor: `exploratory` (default). If the user wants `confirmatory`, create the branch (steps
     2–4) and then route to `govern/preregister` to freeze + register the spec *before* the analysis
     is run — do not execute it as confirmatory until the spec is frozen.
2. **Choose a clear branch name** — `cmp/` (e.g. `cmp/group-diff-y`). The user owns
   naming conventions; suggest one and confirm. Clear, stable names are how comparisons stay
   navigable (the harness does not track branches for you beyond the log).
3. **Create the branch** — delegate to the **datalad doer**:
   > "Create and switch to git branch `cmp/` in this dataset (leave the working tree clean)."
4. **Log it** — append one entry to `project.yaml`:
   `{ ts, op: propose-comparison, stage: analyze, note: "cmp:  ()", branch: cmp/ }`.
5. **Report** — the branch name and that the next step is writing the analysis script, then
   `analyze/run-comparison` to execute it with provenance.

## Constraints
- One comparison = one branch. Do not stack unrelated comparisons on the same branch.
- Keep `project.yaml` append-only.
- Do NOT choose the analysis, model, or figures for the user — the harness scaffolds the edges of
  the work (branch, provenance, log), not the scientific question itself. Writing the analysis
  script is the user's job; `run-comparison` wraps whatever they wrote.
- Delegate the branch operation to the datalad doer; never call git/datalad directly.

## 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-propose-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%.