PyCForge conversion-extension authoring contract

The selected PyCForge installation supplies standardized, inventoried files:
index.json identifies versions and binds each file by SHA-256 and size.
api.xml and api.xsd describe public conversion and authoring methods.
authoring.json follows authoring.schema.json (JSON Schema 2020-12).
workbench.json follows workbench.schema.json and describes generic forms/Help.
Templates and HTML Help are inert data. Unknown optional metadata is preserved;
unknown required protocols must be refused before replacing a compatible cache.

The SDK creates *import extensions using these files and public conversion
services. It does not select, probe or invoke any C compiler, including through
a delegated verification command. PyCForge owns Crucible/C compilation
compatibility and compiler qualification. This file contract contains no
compiler commands, flags, host execution profile or backend-probe instructions.

Authoring workflow: create or recover a project; edit declarations, Help and
fixtures; save explicitly; scan; preview conversion checks; explicitly verify
conversions; package current conversion evidence; register separately.
Verification checks conversion acceptance, expected rejection diagnostics and
source-call coverage. It never runs generated programs. The legacy fixture kind
"run" means a convertible example with preserved author runtime expectations;
stdout, stderr and exit status remain untested declarations. The report and
package provenance identify conversion-only scope. Native/runtime qualification
is a separate PyCForge responsibility and cannot be inferred from SDK readiness.

A 0.x authoring reference may remain browsable by a current SDK, but its native
verification contract is not used. Select an engine supplying authoring format
1.x to verify/package through the compiler-free SDK. Old native reports and
packages stay intact. Ordinary compatible capabilities need no SDK update.
