Best possible solution:

Keep one canonical clean implementation for #80507, land the formatter/test change after acceptable proof, and preserve this PR's redacted CLI output only as supporting context if useful.

Do we have a high-confidence way to reproduce the issue?

Yes. Current-main source shows send and poll dry-runs can reach the formatter as dryRun: true with handledBy: "core", while the formatter only branches on handledBy === "dry-run"; I did not execute the CLI in this read-only review.

Is this the best way to solve the issue?

Yes for the code direction: checking result.dryRun before success formatting is the narrow fix. This PR should not be kept as a second merge vehicle because the same fix is already under review in #80604 and this branch history still contains unrelated removed artifacts.

Security review:

Security review needs attention: The current net diff is clean, but this PR branch history still contains removed workspace/artifact commits that should not be preserved if code is reused.

• [medium] Avoid preserving removed artifact commits — memory/2026-05-07.md:1
  Earlier commits in the PR branch added memory and x-post artifact paths before later removal. A squash or cherry-pick of the net formatter diff avoids publishing those unrelated artifacts into main history.
  Confidence: 0.9

What I checked:

• Current formatter misses the stable dryRun flag: Current main only checks result.handledBy === "dry-run" before send and poll fall through to success-style output, which matches the reported CLI presentation bug. (src/commands/message-format.ts:249, 5ed2a9a4e21b)
• Send and poll carry dryRun separately from handledBy: The current MessageActionRunResult union gives send and poll handledBy: "plugin" | "core" plus a separate dryRun boolean, so the formatter should read dryRun for these variants. (src/infra/outbound/message-action-runner.ts:124, 5ed2a9a4e21b)
• Core dry-run delivery remains skipped: The core send and poll paths return early when params.dryRun is set, so the remaining defect is misleading CLI text rather than real delivery. (src/infra/outbound/message.ts:339, 5ed2a9a4e21b)
• Canonical open PR already carries the same fix: fix(message): show dry-run output for send/poll #80604 is open and changes the same formatter guard to use result.dryRun with send/poll dry-run formatter tests for the same linked issue. (src/commands/message-format.ts:249, 312a25676db7)
• This PR's current net diff is narrow: The live file list for this PR is now only src/commands/message-format.ts and src/commands/message-format.test.ts, with the same central predicate/test change. (src/commands/message-format.ts:249, 37eba6522607)
• Branch history still contains removed workspace artifacts: Earlier commits in this PR branch added memory and x-post artifact paths that later commits removed, so preserving individual commits would still bring unrelated artifact history into main. (memory/2026-05-07.md:1, 4aa4d8e6f8d1)

Likely related people:

• @steipete: Current-main blame covers the formatter guard and result contract lines, and recent GitHub path history shows substantial outbound/message action work on the implicated files. (role: recent area contributor; confidence: high; commits: 5b1d1eac1214, 5e8e77ed83eb, 2ead1502c9bf; files: src/commands/message-format.ts, src/infra/outbound/message-action-runner.ts, src/infra/outbound/message.ts)
• @shakkernerd: Recent path history for src/commands/message-format.ts includes formatter-summary work on the same CLI presentation surface. (role: recent formatter contributor; confidence: medium; commits: 8fe449c8834b; files: src/commands/message-format.ts)
• @vincentkoc: Recent path history shows CLI and terminal formatting work adjacent to the command path that calls the message formatter. (role: adjacent CLI contributor; confidence: medium; commits: be1c38e692f2, 74e7b8d47b18; files: src/commands/message.ts, src/commands/message-format.ts)

Codex review notes: model gpt-5.5, reasoning high; reviewed against 5ed2a9a4e21b.