feat(skills): add ultrareview bundled skill

[emal-avala]

Summary

/ultrareview — review that traces blast radius, not just the diff.

Seven passes:

1. Full diff
2. Callers of every changed public fn (LSP or grep) — check argument/return/error semantic breakage
3. Callees of every changed public fn — find assumptions a new call path might violate
4. Test surface — is there a test for the new behavior? Flag if not
5. Edge cases the diff didn't address (empty/unicode/huge/concurrent/allocator-fail/partial/cancellation)
6. Cross-cutting — logs, metrics, feature flags, migrations, error messages
7. PR body claims vs what the code actually does

Severity-sorted output (critical/high/medium/low) with file:line + remediation.

Anti-slop rule: if the diff is genuinely clean after all seven passes, say so — don't invent findings to justify the review.

Complements /review (fast diff check), /simplify (dead weight), /perf-issue (perf regressions).

Test plan

• cargo fmt --all clean
• cargo check passes
• Manual: run on a real PR and compare with /review — ultrareview should surface caller/test gaps that /review misses

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.