# Preregister

> >

- **Type:** Skill
- **Install:** `agentstack add skill-neuroanalytics-data-science-harness-preregister`
- **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/govern/skills/preregister

## Install

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

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

## About

# Skill: preregister

Turn a comparison into a **confirmatory** one: freeze its specification *before* any data is
analyzed, register that spec with an external registry, and record it as an outstanding
**obligation** in the ledger. The frozen spec is the STAMPED spec-centric research object
(Self-contained, Tracked, Actionable) — the durable thing; the later run is Ephemeral and is checked
*against* this spec. You own the freeze/registration judgment; you delegate the save to the
**datalad doer**.

> This is the confirmatory path of `analyze/propose-comparison`. A "preregistration" recorded *after*
> looking at the outcome is not one — the freeze must precede execution. `analyze/run-comparison`
> then checks results against the registered spec and any deviation is reportable.

## When to use
- A comparison should be confirmatory: its hypothesis, inputs, and analysis plan are to be frozen
  and registered before running (often right after `propose-comparison` picks the confirmatory mode).
- Do NOT use for exploratory quick queries (just `propose-comparison`), or to execute the analysis
  (`run-comparison`), or to resolve/track existing obligations (`govern/obligations`).

## Steps
1. **Freeze the spec** — capture the immutable specification for the comparison's `cmp/`
   branch: `what` (hypothesis), `why`, `inputs`, `expected_outputs`, and the **analysis plan**
   (model, primary outcome, exclusions, corrections). Write it as a durable, versioned file
   (e.g. `code/prereg/.md`) so the frozen text is itself tracked. Confirm with the user that
   this is final — after registration it does not change.
2. **Register externally** — guide the user to submit the frozen spec to the appropriate registry
   (OSF Registrations, ClinicalTrials.gov, PROSPERO for systematic reviews) and obtain the
   registration **id/URL**. The harness records the reference the user provides; it does not
   fabricate or auto-submit one. If registration is pending, record the obligation as `pending`
   with no `ref` yet and note that the ref must be added once assigned.
3. **Record the obligation** — add to `project.yaml` `obligations[]` (per `docs/project-ledger.md`):
   `{ id: prereg-, kind: preregistration, description: "; frozen for cmp/",
   due: , status: pending, ref:  }`.
4. **Log it** — append `{ ts, op: preregister, stage: govern, note: "froze + registered cmp/ ()", branch: cmp/ }`.
5. **Save** — delegate to the **datalad doer**: "save: `datalad save -m 'preregister cmp/: freeze spec + register ()'`."
6. **Report** — the frozen spec path, the registration ref (or that it is pending), the new pending
   obligation, and that `run-comparison` must check outcomes against this spec — deviations are
   reportable. The obligation is discharged (via `govern/obligations`) once the confirmatory
   comparison is completed as specified.

## Constraints
- Freeze precedes execution — never "preregister" a comparison whose outcome has already been seen;
  if it has, record it honestly as exploratory (`propose-comparison`), not confirmatory.
- The frozen spec is immutable after registration — a change means a new comparison + a new
  registration, recorded as such.
- Never fabricate a registration id/URL — record only what the registry actually assigned.
- Add the obligation per the ledger conventions; keep `log:` append-only and the ledger
  schema-valid. Delegate the save to the datalad doer.

## 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-preregister
- 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%.