notes.txt

1) add in datapresser https://datapressr.datahub.io/
2) Look over:  tabpfn-3.5 from priorlabs.ai  POINTLESS
3) maybe a geopandas based skill to work with tabular to spatial format requests.
4) need some prov tracking
5) AI policy document about how we accept the responsibility
6) maybe DOOS needs it's own harness for this?   (and ODSET)

is it all really about the data?

4) is important, also a notebook.md like file would be nice


A not on 3 and 4.   This shows a need for good metadata around the skills in terms of the input and output formats used/produced.  We could look at the workflow hub and see how it is defined there.   However, it would be nice to follow schema.org (SOSO, CDIF, etc) to expresee that more semtntically.

This also speaks to the issue of 3 then to actually document this flow of objects through to the final product.



Nice example:

---
name: database-migration
description: Manages PostgreSQL schema migrations, Drizzle ORM updates, and rollback validation. Trigger whenever files in src/db/ or schema.ts are touched.
---

# Database Migration Skill

## Blocking Pre-Conditions
Before modifying any schema file in src/db/:
1. Run "npm run db:check" to verify the current migration state.
2. Confirm that no pending unapplied migrations exist in src/db/migrations/.

## Execution Rules
- Always use Drizzle Kit generate commands rather than hand-writing raw SQL migrations.
- If altering a non-null column, provide an explicit default value to prevent table locks.
- Never run dropTable without an explicit dry-run confirmation from the user.

## Verification Gate (Mandatory)
Run the following verification sequence in order:
1. Run "npm run db:generate" to generate migration files.
2. Run "npm run test:db" to execute the migration test suite against local Postgres.
3. If exit code != 0, inspect the migration log, revert the change, and apply the Smallest Durable Fix.