This repository is a holding space for Drupal AI ecosystem research snapshots produced with different tool and model combinations.

• main is intentionally lightweight and acts as the landing page for the repository.
• Full research outputs live in branches named after the tool + model combination that produced them.
• Other runs should be published to similarly named branches so the provenance of each research snapshot stays obvious.

Each non-main branch is intended to be a self-contained research snapshot produced by one specific tool/model/reasoning setup.

A branch name should tell you:

• which tool produced the research
• which model was used
• what reasoning mode or depth was used, if that matters

In other words, a branch is not just a working branch. It is a publishable record of one complete run so that outputs can be compared across different tool/model combinations.

• codex-gpt-5.4-high-reasoning: full research corpus produced with Codex using GPT-5.4 at high reasoning

Each branch in this repo is generally trying to answer a version of this question:

"Survey the Drupal AI tooling landscape, especially repos dealing with agent skills, contribution helpers, workflow integrations, prompt libraries, and testing harnesses, then determine what ai_best_practices should absorb, reference, collaborate with, or ignore."

In practice, that usually means the prompt asks the agent to:

• gather and index a defined set of Drupal-related repos and projects
• analyze each repo against the same rubric
• compare overlaps, divergences, and standards conformance across the set
• identify best-in-class patterns, content, and distribution approaches
• produce a synthesis and a draft spec for how ai_best_practices should consolidate the landscape

Research branches should generally contain a structure like this:

• README.md explains the branch at a high level if needed
• research/SPEC.md a proposal or decision draft derived from the research
• research/SYNTHESIS.md a cross-repo or cross-source synthesis
• research/tier3.md lighter scans of adjacent or index-style resources
• research/acquisition-failures.md a record of any missing, failed, or unverifiable inputs
• research/repos/<slug>/INDEX.md deterministic acquisition/indexing output for a specific repo
• research/repos/<slug>/ANALYSIS.md semantic analysis of that repo based on the indexed evidence

Not every branch has to contain exactly the same files, but branches should try to follow this layout so they are easy to compare.

Research branches may include:

• deterministic repo indexes
• per-repo analyses
• cross-repo syntheses
• spec drafts
• acquisition logs and supporting notes

They should not be treated as canonical truth forever. They are time-bound research artefacts, tied to:

• the tool and model that produced them
• the sources available at the time
• the judgment calls made during that specific run

The intent is to compare outputs across tool/model combinations without mixing their artefacts together on main.