Product Release

Use cases that pass in isolation can still fail together: wiring, permissions, and data cross boundaries the isolated tests never touch. A journey test that runs verified use cases to one observable outcome through the deployed public boundary proves the connected path customers actually take.

Reviewed forWhich completed journeys lack a passing test through the deployed boundary? Look at the journey tests beside each included journey.

Marked guidance: a reviewer answers this question.

Each boundary carries failures the others cannot see: frontend checks say nothing about migrations, and backend suites say nothing about rendered screens. Running every check the changed boundaries select keeps each change evidenced where its failure lives. A narrow category that passes alone certifies only part of the change.

Reviewed forWhich changes lack a passing result for a check their boundaries select? Look at the change report beside the boundary table.

Marked guidance: a reviewer answers this question.

The release build compiles the solution in release mode, catching the conditional and optimization-gated breaks a debug build hides. The architecture, integration, and acceptance suites the size requires prove architecture layering, wiring, and behavior in turn. Skipping any of them ships backend code the project's own suites do not fully prove.

Reviewed forWhich backend changes lack a passing release build or suite result? Look at the build and suite reports beside the change.

Marked guidance: a reviewer answers this question.

Each step guards a different failure: frozen installation pins dependencies, lint and type checking catch static defects, tests prove behavior, and the production build proves the shipped bundle compiles. A frontend that skips one arrives at release with that whole class of defect unchecked.

Reviewed forWhich frontend changes lack a passing install, lint, type, test, or build result? Look at the frontend reports beside the change.

Marked guidance: a reviewer answers this question.

A journey change alters the connected path, so only a journey test through the deployed public boundary can prove it still reaches its outcome. Narrower suites keep passing while the journey itself breaks.

Reviewed forWhich journey changes lack passing journey tests? Look at the journey reports against the deployed boundary.

Marked guidance: a reviewer answers this question.

G checks read records only: shape, references, links, reachability, coverage, and the Behavior rules the check list names. Without them, a renamed record or a moved file silently breaks the references other pages rely on, and the book renders a graph that no longer holds. litespec model check --since <ref> blocks only on findings about records and code changed since <ref> and reports the rest, so an adopting project clears new findings first.

Checked bylitespec model check exits zero with no G finding.

Marked checkable: the tool named here checks it.

An extension adds behavior, tests, and operating surface beyond the baseline. Its stated verification is the only evidence written for that surface, so skipping it releases extended behavior the baseline checks never examine.

Reviewed forWhich affected extensions lack their stated verification evidence? Look at the extension evidence beside the change.

Marked guidance: a reviewer answers this question.

OpenAPI and frontend API types derive from a source contract. A source change that leaves the committed copies stale ships clients built against a contract the server no longer honors.

Reviewed forWhich committed contracts differ from a fresh generation? Look at the contract diff beside the source change.

Marked guidance: a reviewer answers this question.

Timestamps, machine paths, and unstable ordering make every regeneration look like a behavior change. Deterministic output keeps contract diffs reviewable, so a reviewer sees only the changes that matter.

Reviewed forWhich contract diffs hold timestamps, machine paths, or unstable order? Look at the generated diff beside each contract.

Marked guidance: a reviewer answers this question.

An unqualified reader blocks on noise or passes on silence, and each mistake trains the team to ignore its findings. Passing a known good tree and reporting each planted defect in a known bad tree proves the reader sees what it claims before it blocks any change.

The .NET reader passes the clean tree and reports the planted undeclared write before its S findings block.

Reviewed forWhich gating S readers lack passing fixture runs? Look at the fixture results beside the stack reader.

Marked guidance: a reviewer answers this question.

Restricted behavior without authentication accepts callers nobody identified, and without authorization it accepts actions nobody permitted. Both decisions call for a trusted identity and a declared permission before the behavior ships.

Reviewed forWhich restricted release paths lack authentication or authorization? Look at the guards on each restricted path.

Marked guidance: a reviewer answers this question.

Evidence scattered across builds and reports answers no single question. One release record per immutable artifact binds every result to the exact output it proves.

Reviewed forWhich releases lack one record naming the immutable artifact? Look at the release records beside each artifact.

Marked guidance: a reviewer answers this question.

An expired exception is a promise past its date, and an unverified external claim behind released behavior is a promise with unsettled terms: a provider fact, a law text, or a term nobody rechecked. Settling each one, or excluding the affected behavior from every owning record, keeps the Intent and release records agreeing on what shipped. exception-expired reports every expired exception.

Reviewed forWhich releases depend on an expired exception or an external claim without excluding the behavior? Look at the exceptions and claims beside the owning records.

Marked guidance: a reviewer answers this question.

A vague report lets unrun checks hide among completed ones. Naming exact commands, outcomes, scope, omissions, and limits tells a reviewer precisely what the evidence covers and what it does not.

Reviewed forWhich completion reports lack commands, outcomes, scope, omissions, differences, limits, or covered use cases? Look at the report beside the change.

Marked guidance: a reviewer answers this question.

The source commit names the verified input and the artifact reference names the deployed output. Together they bind the record to one build, so evidence from another build cannot stand in for it.

Reviewed forWhich release records lack a source commit or artifact reference? Look at the record header beside each artifact.

Marked guidance: a reviewer answers this question.

The behavior boundary of an artifact is exactly its journeys, use cases, and extensions. Listing them tells operators and reviewers what the artifact contains and, by omission, what it does not.

Reviewed forWhich records omit an included journey, use case, or extension? Look at the behavior list on each release record.

Marked guidance: a reviewer answers this question.

Specifications own behavior and scenarios. A release record that restates them becomes a second copy, and the copies drift until nobody knows which one holds. The record links outward instead.

Reviewed forWhich release records restate a specification or copy a scenario? Look at the record beside the linked specifications.

Marked guidance: a reviewer answers this question.

Accepted debt with no expiry date is forgotten work. An exception of kind gap with one turns it into a dated promise whose findings report until the date and whose exception-expired reports after it. The model page states the exception shape.

Claimwick ships month-end export without its archive proof, so the debt lives as an exception of kind gap with an expiresOn of 2027-03-01; exception-expired reports it after that date.

Reviewed forWhich accepted debts lack an exception of kind gap with an expiry date? Look at the exceptions beside each debt.

Marked guidance: a reviewer answers this question.

A release judged from fragments is judged by archaeology: the reader digs through other records to fill the gaps. A complete record lets one reader judge one release from one page, with no upgrade guide filling in.

Reviewed forWhich release records depend on an earlier record to be judged? Look at the record alone beside its links.

Marked guidance: a reviewer answers this question.