docs: names reference — how Compose and Module were chosen

[wmadden-electric]

Adds agent-os/product/names-compose-and-module.md — the consolidated reference for the two settled names, written for anyone who wasn't in the room:

• The names at a glance — Prisma Compose (framework), Module (unit), Prisma App (artifact), plus the supporting registers, and the one-breath sentence that uses them all.
• The philosophy — three principles, each learned by shipping a wrong name first: products are named for value but names must refer as well as describe; vocabulary nouns are tested in speech, not on the page; identity belongs to the registry, not the unit noun.
• The rubrics — the product value tests, the referential battery (workbench frame, artifact-collision hard rule, bare token, identity frames), and the eight unit-noun spoken tests.
• Why Module / why Compose — the Hex → System → Module and MakerKit → Prisma App → Compose histories, what each failure taught, why the winners pass, and the runners-up with their disqualifiers.
• Lessons worth keeping — the transferable rules (test in speech; test at the smallest grain; never name a tool after its output; the gloss test; verify "nobody says X").

Linked from the product README as the naming entry point; ADRs 0025/0026/0027 and vocabulary-tests.md remain the binding records.

🤖 Generated with Claude Code

[wmadden-electric]

Closing as superseded by the Composer rename (8d495c6, feat!: rename framework Prisma Compose to Prisma Composer).

This doc's headline claim — the framework is Prisma Compose — is no longer the settled name: "Compose" collided audibly with Compute and read as a bare verb, so the framework became Prisma Composer. The rename also rewrote the docs this PR was consolidating: agent-os/product/naming.md now carries the full rationale (including why compose beat the field and why the agent-noun form won), and vocabulary-tests.md records the rubrics and outcomes with the Composer result. The methodology and history this doc captured are preserved there and in ADRs 0025–0027.

A fresh "how the names were chosen" retrospective would need to be written against the Composer outcome rather than rebased from this branch.