You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Added
The Stage rendering skill teaches case tables (one generated specification class per case, named <Spec>_<Case>, from Stage 4.59.0; a name collision is refused with STAGE-CRATIS-004) and whole numbers rendered as long from ESM v8 (Stage 4.61.0; below v8 an out-of-int literal is refused with STAGE-ESM-031), and states that the CLI 3.46.1 default renderer (Stage 4.52.0) applies neither, with one evaluation per rule (#591)
The model-authoring and model-review skills teach semantic-diffbefore/after sources from Screenplay 4.131.0: the active workspace, a model folder or registered worktree, or an export, so a model is compared with main through a worktree path instead of two exports, and when to use read-proposal with view: "semantic-diff" instead (#592)
The cratis-arc-command and cratis-specifications-csharp skills teach Arc generated values: [GeneratedValue] on Handle() parameters for server-owned identifiers, when not to generate, analyzers ARC0022–ARC0025, and pinning generated values in CommandScenario specs with PinGeneratedValue, AllowGeneration and GeneratedValues (part of #576)
Changed
The pinned standalone Screenplay tool and published contract move to 4.131.0, and version statements across the Screenplay skills separate the current pins (CLI 3.46.1 bundling Screenplay 4.130.0 and Stage 4.52.0) from historical probes (#592)
A skill may carry three to five evaluation tasks instead of exactly three (part of #591)