Character / CHARAL 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-character/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 capture-first ground truth, delta-first scoping, known-open register (verify, never re-file), deduplication, 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 route per the canonical's cross-audit routing: component shape to /audit-ecs, AVIF/CLAS/NPC_ parsing to /audit-esm, CTDA evaluation to /audit-scripting, scheduler access to /audit-concurrency, gameplay writers to /audit-gameplay, and ActorValues save schema to /audit-save.
• Read the charal-* capture documents before any Rust constant; a numeric finding without a capture-line Source is not reportable.
• Prefer the repository's codebase-memory graph tools for ruleset construction sites, derived-stat consumers, and population-boundary call chains; use text search for GameKind branches, roster FormIDs, capture-document terms, 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.
• 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.