Audit Suite Orchestrator

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

Load the Canonical Command

1. Read .claude/commands/audit-suite/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 three modes (--preset, --area, --changed routed through _audit-owners.md and _audit-route.sh), its launch rules including the orchestration hazard (synchronous dimension analysis or /tmp/audit/<name>/dim_N.md scratch files, verified against each written report), report verification, and the summary-table merge 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 dispatch each routed audit by reading its owner SKILL.md, adding the owner's neighbor audits exactly as the canonical tables define them.
• Derive audit sets with git ls-files / git diff --name-only piped through _audit-route.sh; use codebase-memory or text search only to resolve an unclear owner or path.
• Interpret Claude background Task fan-out as Codex sub-agent delegation — the suite's normal mode, max 3 concurrent; without delegation, run the routed audits sequentially and merge their reports as usual.
• Preserve each routed audit's canonical output path and do not create GitHub issues; publishing remains a separate audit-publish action and never belongs in a suite.

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