Regression Verification 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-regression/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 churn-weighted issue discovery, fix and guard-test location steps, PASS/PARTIAL/FAIL/UNVERIFIABLE status rules, unconditional fragile-area guards, 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 fragile-area guard's deep checklist lives there.
• Prefer the repository's codebase-memory graph tools for locating fix symbols, fix sites, and guard tests; use text search for issue-number citations in tests and source, Fix #N commit subjects, and graph gaps.
• Interpret Claude Task-agent instructions as Codex sub-agent delegation only when delegation is available and allowed. Otherwise work the fix → guard-test chain one issue at a time.
• Reading existing GitHub issues via gh stays in scope for discovery and verification; creating issues remains reserved for a separate audit-publish action. Preserve the canonical output path.

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