Implementation and evidence.
| Field | Context |
|---|---|
| Implementation | Source-reviewed internal development |
| Evidence/readiness | Package qualification pending |
| Usability | No numeric or named assessment assigned |
| Version/platform | Internal foundation context; no selected public package or rated environment |
| Resolve/client | No release-supported edition/transport/client assignment here |
| Review basis | Curated source-reviewed development descriptions; not a fresh runtime walkthrough |
A scenario worth assessing.
Create a90-frame comp, mark a subset and inspect the selected output range.
Define the starting state, intended result and environment before assessing. Preserve actual observations, including refusal, partial result or inconclusive output. Do not choose a score first and then search for supporting proof.
Human workflow profile.
| Dimension | Assessment |
|---|---|
| Discoverability | Not assessed |
| Creation and editing | Not assessed |
| Feedback | Not assessed |
| Recovery | Not assessed |
Observe the actual user route: find the relevant controls, complete the scenario, understand the visible result and recover appropriately. Headless document changes do not establish human discoverability or feedback.
MCP agent workflow profile.
| Dimension | Assessment |
|---|---|
| Discovery | Not assessed |
| Task completion | Not assessed |
| Inspection | Not assessed |
| Revision | Not assessed |
| Recovery | Not assessed |
Use a named client to discover state, complete a bounded task, inspect the result, revise after feedback and handle stale state/disconnects. A command catalogue entry does not prove task completion.
Fusion delivery profile.
| Dimension | Assessment |
|---|---|
| Appearance | Not assessed |
| Animation | Not assessed |
| Editability | Not assessed |
| Transfer reliability | Not assessed |
Inspect actual destination appearance, motion, a supported editable change and reliable target/transfer/persistence. Compiler structure alone cannot establish this profile.
Qualification and revalidation.
- Choose the exact candidate/package and environment.
- Complete relevant human/client/output scenarios and inspect results.
- Retain limitations and actual failure/recovery evidence.
- Clear public evidence of private information before publication.
- Revalidate affected proof when implementation, resources or the environment change.