Physics / PHYSAL 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-physics/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 dimension order, ground-truth doc reading, guard verification, known-open register checks, and 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 crosses the canonical handoff boundaries (gameplay behaviour, WATR translation, water shading, bhk* parsing, CollisionShape extraction, lock order).
• Prefer the repository's codebase-memory graph tools for callers, phase-order traces, and lock-order analysis; use text search for guard symbols, constants, GameKind/bsver seams, configs, docs, and graph gaps.
• 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 and closed-fix guards as regression baselines: verify they still hold, never re-file them as new.
• 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.