Gameplay Audit

Use the Claude Code command as the canonical workflow instead of maintaining a duplicated snapshot.

Load the Canonical Command

1. Read .claude/commands/audit-gameplay/SKILL.md completely.
2. Read .claude/commands/_audit-common.md and .claude/commands/_audit-severity.md completely before auditing.
3. Follow the canonical command's delta-first dimension selection, per-dimension paths and guards, ground-truth doc and smoke-gate re-verification, known-open register, deduplication, and invariant-matrix report format.

Adapt Claude Conventions to Codex

• Treat the user's text after the skill request as the command arguments described by the canonical skill's argument-hint.
• Resolve each /audit-<name> reference to that audit's SKILL.md under .claude/commands/, and load a sibling audit only when a finding lands outside the gameplay layer (CHARAL formulas → character, save schema → save, lock order → ecs + concurrency, HUD rendering → ui, KCC/ragdoll → physics, ALCH/PACK decoders → esm, Papyrus → scripting); this skill owns what gameplay state should be restored.
• Prefer the repository's codebase-memory graph tools for caller tracing, per-actor runtime-component inventories, and written-never-read checks; use text search for stage-order comments, save_io registry rows, guard test names, and determinism greps.
• Interpret Claude Task-agent instructions as Codex sub-agent delegation only when delegation is available and allowed. Otherwise audit one dimension at a time.
• Treat the canonical known-open register as cite-don't-refile context, and never launch a second engine beside the user's.
• Preserve the canonical output path and do not create GitHub issues; publishing remains a separate audit-publish action.

Do not copy the canonical command body into this skill. The Claude command is the single source of truth.