aries-writeback

ARIES Writeback

What you're doing

The user has forecasts in Crude Code terms — usually the curves asserted in this session's deal_forecast_wells run (their own independent forecast, or a revised take on a seller's deck) — and wants them in ARIES. This skill packages them as an import package: a zip of CSVs the target database's engineer appends with ordinary MS Access import, under a new qualifier so nothing existing is touched. Think of it as the reverse of aries-to-valuation, with the same pinned conventions run backwards (nominal-monthly → effective-annual).

This is v0 by design: no binary is created or modified. The deliverable is rows plus instructions; the human does the import on a copy of their own file. Say that plainly when handing it over.

What you need

• The curves, from this session. You asserted them (or translated them) — you already hold {qi, di, b} per stream, the anchor, and the wells. Do not re-derive parameters from anywhere else.
• Code execution for the packaging script.
• The target database's _aries/ dir when available (from the explorer's triage): it fills each well's PROPNUM by API and guards against reusing an existing qualifier. Without it, PROPNUMs the user can't supply ship as placeholders with join-on-API instructions.

Workflow

1. Write curves.json (schema in aries_package.py's docstring): qualifier (short, NEW — e.g. CC2608), and per well: api, anchor_month, oil/gas params exactly as committed via deal_forecast_wells, optional cums and propnum. Copy numbers verbatim — never round, never adjust.
2. Build the package:

   python3 aries_package.py curves.json --aries-dir _aries
   

   (--aries-dir whenever the target database was triaged this session.) Read the printed summary and notes.
3. Hand the zip to the user as a downloadable file, with a two-line explanation: it adds a new qualifier alongside their existing scenarios; the README inside walks their engineer through the Access append on a copy. Relay any PROPNUM-placeholder notes explicitly.
4. If the session also produced a valuation, remind the user which run the exported curves came from — the package and the deal sheet should tell the same story.