Skip to content

Using These Docs with AI Agents

If you use an AI assistant to prepare a submission, give it the resources below rather than letting it guess. They contain the same rules as these pages, in a form an agent can follow.

Agent skill

Download the ImmPort data submission skill (zip)

The skill is a folder with a SKILL.md file, which is the open format several AI coding assistants use for skills. It teaches an assistant the submission workflow, the rules it must not break (including de-identification), how to pick templates and package them, how to read validation errors, and how to call the upload API.

  • Claude Code and other tools that support SKILL.md: unzip it into your skills folder (for Claude Code, ~/.claude/skills/).
  • Claude apps: upload the zip as a custom skill in Settings.

The skill is designed to work with an ImmPort submission MCP server when one is available, and with the REST API until then.

Machine-readable data

Resource What it contains
submission-data.json All templates with purpose, category, assays and download links; the upload workflow for every assay with its same-package groups; the package loading order; the rules (R-01 to R-24); and the error messages (E-01 to E-33) with causes and fixes
templates.json Every column of every template: required flags, types, lengths, lookup table, foreign keys, JSON attribute names
lkTables.json The allowed values for every controlled vocabulary field
/llms.txt A short index of these pages for language models, following the llms.txt convention

submission-data.json is generated from the tables on these pages: the Template Catalog, Assay Workflows, Rules at a Glance and Troubleshooting Errors. The pages and the data always say the same thing.

Conventions on these pages

  • Stable IDs: rules are R-xx and errors are E-xx. Each one is also an anchor, for example rules-at-a-glance/#R-08 or 03-validation/troubleshooting-errors/#E-18, so you can cite or link a specific rule or error.
  • Structured tables: catalog, workflow, rule and error rows carry data-* attributes (data-template, data-assay, data-rule, data-error) for reliable parsing.
  • Text alongside diagrams: every diagram is backed by text or a table on the same page.

Privacy

Don't paste identifiable human subject data into a general-purpose AI service. Work with de-identified data, or use a tool your institution has approved for that data (R-19).