← ClaudeAtlas

game-studio-factorylisted

Operate the whole game from an Accepted Playable Baseline with explicit acceptance and regression. Resolve project state first; preserve existing CLI and historical formats.
csonic0926/game_studio_factory · ★ 0 · AI & Automation · score 60
Install: claude install-skill csonic0926/game_studio_factory
# Studio Factory operation Resolve GAME_REPO from explicit path or current Git root. Resolve STUDIO_ROOT from `design/STUDIO_FACTORY.local.md`, legacy `design/AI_FACTORY.local.md`, installed manifest or this skill's real source path. If unlinked, invoke `init-game-studio-factory`. Never hardcode a developer path or scan sibling games. 1. Run `python3 $STUDIO_ROOT/factory.py inspect --game-repo $GAME_REPO`. 2. MIGRATION_REQUIRED: preview `migrate --check --project-id <id>`; apply only an explicitly authorized migration with its exact source digest. No automatic approval/history conversion. Legacy CLI calls/parameters remain supported. 3. For v2 run `factory.py context --game-repo $GAME_REPO --capability studio --task objective` with the actual task kind and current task id when known. 4. Follow `$STUDIO_ROOT/factory_core/docs/WORKFLOW.md` and the returned capability rules. Continue in this primary agent; load task methods only as needed. Freeze one full design, run the two independent reviews on its exact hash, obtain required USER rulings, produce within scope, then exact-output evidence and specialist acceptance. Ordinary replies do not create reviews. 5. Checkpoint validated references, progress and unresolved questions. A resumed session rechecks the ledger; context views and technical success are not authority or whole-game delivery. Whole-game delivery still requires new gameplay acceptance, exact-build USER playtest, predecessor regression