Problem
Codex is valuable not just for reviewing diffs but for verifying other artifacts that Claude produces: implementation plans, task trackers, and Claude's own output. Currently there's no structured way to ask Codex to verify these.
Proposal
Add verification commands:
/codex:verify-plan
- Reads the active plan or handoff document
- Asks Codex to check: feasibility, missing steps/dependencies, risk areas, operation ordering, alternative approaches
- Returns structured verdict (PASS/CONCERNS/FAIL)
/codex:verify-output
- Captures Claude's most recent substantive output
- Asks Codex to independently verify correctness, completeness, and missed edge cases
- Cross-model verification: catches blind spots from a different model's perspective
/codex:verify-tracker (more niche, could be deferred)
- Reads implementation tracker files (YAML/markdown)
- Verifies: task completeness, metrics consistency, phase dependencies, blocker tracking
Why this matters
Cross-model verification is one of the strongest use cases for having Codex alongside Claude. If Claude produces a plan, having Codex independently check it catches assumptions and blind spots that same-model self-review misses.
Problem
Codex is valuable not just for reviewing diffs but for verifying other artifacts that Claude produces: implementation plans, task trackers, and Claude's own output. Currently there's no structured way to ask Codex to verify these.
Proposal
Add verification commands:
/codex:verify-plan/codex:verify-output/codex:verify-tracker(more niche, could be deferred)Why this matters
Cross-model verification is one of the strongest use cases for having Codex alongside Claude. If Claude produces a plan, having Codex independently check it catches assumptions and blind spots that same-model self-review misses.