Source: https://github.com/aipoch/medical-research-skills

LabArchives Integration

This skill is an Academic Writing workflow built around LabArchives evidence. Its goal is not just API access, but turning authorized ELN material into manuscript-ready writing deliverables.

When to Use

• The user has LabArchives notebook content and needs a Methods draft grounded in recorded procedures.
• The user needs a data availability statement, reproducibility appendix, experiment timeline, or submission support summary based on ELN records.
• The user wants to gather authorized notebook data first, then convert it into academic writing outputs.
• The user needs a deterministic evidence-to-writing workflow instead of freeform summarization.

When Not to Use

• The user asks for unauthorized access to notebooks or other users' data.
• The user wants clinical recommendations, diagnosis, or treatment language.
• The user asks you to fabricate notebook records, timestamps, protocol details, or compliance statements that are not present in the source.
• The user has no authorized export, no notebook metadata, and no textual record to ground the writing output.

Primary Writing Outputs

This skill supports these deliverables:

• Methods Draft Based on notebook entries, protocols, instrument logs, and sample-processing notes
• Data Availability Statement Based on notebook identifiers, repository links, export status, and sharing constraints
• Reproducibility Appendix Based on protocol versions, software environments, parameter logs, and file provenance
• Experiment Timeline Summary Based on dated entries, milestones, and decision points
• Submission Support Note Based on notebook scope, audit trail, and documentation completeness

Authorized Input Sources

Use one or more of:

• exported notebook text or JSON
• manually pasted LabArchives entry content
• protocol summaries
• experiment metadata tables
• authorized backup output from bundled scripts

Optional collection step:

• scripts/setup_config.py
• scripts/notebook_operations.py
• scripts/entry_operations.py

Writing Output Contract

Output A: Methods Draft

Must include:

• study material or sample context
• experimental workflow in chronological order
• instrument / assay / software mentions if present in source
• quality-control or versioning note if present in source
• no invented parameter values

Output B: Data Availability Statement

Must include:

• what data are available
• where they are stored or how they can be requested
• any access restrictions
• relationship to LabArchives or downstream repository

Output C: Reproducibility Appendix

Must include:

• protocol version references
• software or pipeline identifiers if present
• provenance or notebook traceability note
• missing-record warning if the audit trail is incomplete

Output D: Experiment Timeline Summary

Must include:

• dated milestone order
• major protocol transitions
• validation / repeat / deviation points if documented

Workflow

1. Validate authorization and source sufficiency

Confirm:

• the requester has authorized access to the notebook data
• the source contains enough grounded information for the requested writing output

If not, stop and use the refusal template in ## Refusal and Recovery Contract.

2. Choose the acquisition path

Use direct source text if already available. Prefer this path for speed.

If data must be collected first, use one of the bundled scripts:

python scripts/setup_config.py
python scripts/notebook_operations.py --help
python scripts/entry_operations.py --help

Use --dry-run where available before live execution.

3. Normalize notebook evidence

Extract only writing-relevant elements:

• dates
• protocol names and versions
• sample or cohort descriptors
• software / pipeline names
• QC notes
• repository / export details
• compliance or sharing constraints

4. Draft the requested academic writing output

Keep the prose:

• factual
• audit-trail grounded
• publication appropriate
• free of operational noise that does not belong in the manuscript deliverable

5. Run the final writing safety pass

Check that:

• every claim maps back to source evidence
• missing evidence is labeled as missing
• no compliance statement is invented
• no unauthorized identifiers are surfaced

Refusal and Recovery Contract

If the workflow cannot proceed safely, respond with:

Cannot generate the requested LabArchives-based writing output yet.
Reason: <missing authorization / insufficient export / incomplete metadata / unsupported request>
Minimum next step:
- <step 1>
- <step 2>

Use this contract for:

• missing notebook authorization
• no usable export or pasted content
• requests to infer missing protocol details
• requests to expose restricted data

Script Usage Notes

The bundled scripts are supporting collection utilities, not the final output themselves.

• setup_config.py: create or validate configuration
• notebook_operations.py: list notebooks, plan backups, or perform authorized exports
• entry_operations.py: inspect entry-level content or upload artifacts when explicitly needed

If a script path fails:

• report the exact command
• report the exact failure
• continue with direct writing only if enough grounded notebook text is already available

Academic Writing Style Rules

• write in neutral, methods-oriented academic prose
• prefer verifiable chronology over interpretive narrative
• do not overclaim documentation quality if the notebook trail is partial
• clearly distinguish documented, not documented, and not provided

Recommended Templates

Use assets/writing_outputs_template.md as the default skeleton for the four main writing deliverables.

Deterministic Rules

• keep output headings stable
• do not expose raw credentials, tokens, or private notebook identifiers unless the user explicitly needs authorized internal formatting
• if a timestamp or version is missing, say it is not documented
• treat data availability and reproducibility statements as formal manuscript components, not casual notes

Completion Checklist

• Authorization boundary checked
• Source sufficiency checked
• Requested writing deliverable selected explicitly
• All statements grounded in notebook evidence
• Missing evidence labeled rather than invented