STAGE 0.5.14

Released: 2026-07-23

STAGE 0.5.14 incorporates the failed detached visual-rehearsal experiment as a workflow correction. Visual work stays inside ordinary project delivery and uses the real engine, game, content, and production consumer. Persistent galleries or almanacs are normal project features only when they have value for players, authors, or recurring developers independent of STAGE evidence.

Why This Patch Exists

The detached rehearsal path accumulated screenshots, videos, contact sheets, HTML, configuration, and review records. Most of that machinery was technically functional but did not answer useful product questions. It duplicated the engine, rendered placeholders instead of canonical content, and made the evidence package its own consumer.

The useful exception was a motion preview of Enemy Arrival. It worked because it showed real production content over time and answered a concrete question. That lesson belongs in ordinary visual delivery; it does not justify a parallel review product.

Changes

• Keep visual implementation, diagnosis, audits, and polish in stage-deliver-change.
• Keep the native production path, human visual authority, and prohibition on parallel review products in the always-loaded skill body.
• Move claim-specific visual detail into the conditional references/visual-work.md reference.
• Prefer live engine or game inspection when the owner can inspect it directly.
• Treat captures and portable pages as exports only when a named remote, comparison, CI, audit, release, or retention consumer needs them.
• Treat a codex, gallery, bestiary, model viewer, replay browser, inspect mode, or in-game almanac as an ordinary backlog feature with its own user and acceptance criteria.
• Document how skill references may reduce default burden without hiding universal rules.

Practitioner Impact

An ordinary nonvisual change no longer loads the detailed visual policy. A material visual claim follows one link at the point where camera, lighting, animation, VFX, UI, procedural readability, or other rendered context becomes relevant.

The first question is now which native production view can answer the claim. STAGE does not ask the project to create a contact sheet, capture plan, review host, or evidence package. If the game genuinely benefits from a gallery or almanac, build and verify that feature for its actual user through the normal project workflow.

Compatibility

The plugin package version is 0.5.14; SPEC.md is method version 0.5. Project-map and case-study artifact versions 0.1 and 0.3 remain supported. No game-project migration is required.

Verification

The release is checked through:

• the complete deterministic STAGE release gate;
• direct Codex plugin and both skill-package validators;
• workflow-routing tests that resolve the conditional reference together with the primary skill;
• repository-local project-map, case-study, schema, and Markdown-link validation; and
• source-to-installed-package parity checks after installation.

No new model-bound behavior probe is claimed for this patch. Static package validation can establish the progressive-disclosure layout and policy boundaries, but not prove when a model will choose to read the reference or the quality of a human visual judgment.

Limitations

This correction is derived primarily from The Circussy One visual-rehearsal dogfood and the owner's direct rejection of its utility. It supports a strong default toward native, production-consumer inspection. It does not establish that detached capture is never useful, that every game needs an almanac, or that live engine inspection is sufficient when remote review, comparison, CI, audit, release, or retention is the actual consumer.