Phase 7: artifact contracts + completion/correctness truth (deferred)

[macanderson]

Phase 7 — artifact contracts + completion/correctness truth. (Deferred — after the goal is reached.)

Deterministic contract validators + the brand-kit witness (a large fixture). Completion/correctness truth for produced artifacts. High value but independent of the learning loop — do when completion-truth becomes a priority.

Blocked by Phase 1 installment 3 (contract + outcome types). Part of Epic #469.

[macanderson]

Triage 2026-07-26 — the pure validators already shipped; the residual is execution

More of this phase landed than the body reflects, because #578 pulled work back from "the repository's job" into the pure layer.

Shipped. stella-core/src/context_record/contract.rs (649 lines) now enforces the §8.13 coverage rule whole: ContractValidation::validate_against(&contract) at :337 (duplicate/unknown/missing result ids → error), and non_executable_required_ids() at :212 filtering r.required && !r.requirement_kind.is_recognized() so an unknown required kind fails closed (:384). outcome.rs (286 lines) links the dimensions: error and failed both block a complete verdict; skipped deliberately does not.

Not shipped — the executor. The code says so at contract.rs:28:

an unrecognized kind is non-executable and fails closed when required (enforced by the executor in a later phase).

The ten recognized kinds (FILE_EXISTS, DIRECTORY_EXISTS, GLOB_MIN_COUNT, … contract.rs:36+) are string constants — nothing ever stats a path or expands a glob.

The brand-kit witness is not the fixture this issue asks for. It exists only as unit-test scaffolding: contract.rs:399 name: "brand-kit", :409 task_intents: ["create_brand_kit"], context_use.rs:657 — a few dozen lines, not a large end-to-end fixture.

Suggested rescope: retitle to "Phase 7: artifact-contract executor + brand-kit witness fixture" and note in the body that the deterministic pure validators and the completion↔contract link shipped in #578. Residual: the requirement executor (filesystem/argv/semantic-judge methods), the real fixture, and wiring OutcomeAssessment to actually-produced artifacts — all of which still need Phase 2/3 for identity resolution.

[macanderson]

Superseded by the four-phase adaptive-context plan merged in #692 (docs/design/adaptive-context-plan.md), which replaces the eleven-phase roadmap and the 2026-07 spec bundle this issue was written against.

Disposition — old Phase 7 (artifact contracts + completion/correctness truth): deferred, named explicitly in spec §9 so it is a decision rather than an omission.

Valuable and independent of this loop, but large. The type layer already exists (#471) and nothing consumes it. Revisit when completion truth is the priority.

Closing as superseded rather than abandoned — see the old→new table in the plan's Tracker reconciliation section. Epic #469 has been rewritten around the four phases.