Open science practices

Open science extends what this pack does for software to the whole research workflow: materials, data, analysis plans, manuscripts and reviews, open by default and closed only with reasons. For an agent the job is concrete plumbing, not advocacy: set up the platforms, wire the links between artifacts, and make each openness claim verifiable. The payoff structure is worth stating when relevant: preregistration protects claims from hindsight, preprints establish priority and speed feedback, open materials earn reuse and citations - openness is self-interest aligned with integrity.

The OSF project as the hub

The Open Science Framework (osf.io, run by the Center for Open Science) is the general-purpose spine for open projects:

• Structure: one OSF project per study, with components for materials, data, analysis code and preregistration; connect the code repository rather than copying it (OSF links forges; the repository stays the working home - rseng-version-control-review).
• Every component gets a persistent identifier and citation; wire these into the paper and the repository README so artifacts reference each other both ways (the same both-ways discipline as rseng-archiving).
• Access control is per-component: public materials next to embargoed data next to private drafts is a normal shape - "as open as possible, as closed as necessary", with reasons stated (rseng-regulatory-compliance governs the closed parts).
• OSF has a REST API; bulk setup and link auditing are scriptable when a lab runs many projects.

Preregistration

A preregistration is a frozen, timestamped analysis plan filed BEFORE the data is seen - the strongest cheap defense against hindsight bias, and increasingly expected in several fields:

• File it on OSF's registration system (or a domain registry); once registered it is immutable - that is the point.
• The plan should be code-shaped where possible: hypotheses, exact variables, exclusion rules, models and inference criteria
  • ideally with the analysis script itself attached, run on simulated or pilot data (rseng-reproducibility's discipline applied before data collection; power-analysis code belongs here too).
• Deviations happen legitimately: the practice is to REPORT them as deviations with reasons, not to hide them - preregistration separates confirmatory from exploratory, it does not forbid exploration.
• Registered Reports go further: peer review of the plan before results exist, with in-principle acceptance - suggest the format when the field's journals offer it.

Preprints

• Deposit at submission time (arXiv-class or domain servers; ASAPbio documents the landscape and journal policies) - a DOI'd preprint establishes priority and makes the work citable months or years early.
• Check the target journal's preprint policy first (most allow it; a few constrain versions) and use the journal's own checking where unsure.
• Link the chain: preprint to published version (servers support the link; OpenAlex tracks it - rseng-citation-hygiene uses this when verifying references), and both to the OSF project, data and code DOIs. The linked cluster IS the open research object.
• Versioned preprints are normal; update after major revisions so readers land on current claims.

Open peer review and open participation

• When venues offer open review, the practices from rseng-software-peer-review transfer: constructive, specific, evidence-based - with a name on it.
• Review artifacts too: data and code review during peer review (CODECHECK-style execution) is open science applied to the evidence layer (rseng-software-peer-review, rseng-research-integrity).
• Community review of preprints (public comments, review platforms) counts as scholarly contribution; encourage recording it (ORCID records review activity).

Keeping openness claims honest

Before submission or reporting, audit the claims: "data available at X" resolves and matches (rseng-archiving's verify step), "preregistered" links to the actual registration and deviations are reported, "code available" points at a tagged, licensed, citable release (rseng-publishing-releasing, rseng-licensing, rseng-citation-metadata). The open-science story is the sum of verifiable links - and AI contributions to any of it are declared in aidecl.yaml (rseng-ai-declaration).

Working with this skill

This skill is source-independent: its authority is the OSF and Center for Open Science documentation and ASAPbio resources linked below. It is the researcher-workflow face of what rseng-fair-software, rseng-reproducibility and rseng-archiving provide at the artifact level.

Learn more (verified):

• https://osf.io - Open Science Framework
• https://www.cos.io - Center for Open Science
• https://www.cos.io/initiatives/prereg - preregistration guidance
• https://asapbio.org - ASAPbio preprint resources
• https://orcid.org - ORCID researcher identifiers

Related skills

Check whether any of these applies before moving on:

• rseng-archiving - both-ways artifact linking discipline
• rseng-citation-hygiene - preprint-to-published version drift
• rseng-publishing-releasing - citable code release behind claims
• rseng-regulatory-compliance - governs the closed components
• rseng-research-integrity - preregistration deviations reported honestly
• rseng-software-peer-review - open review practices transfer