Refuse to move a bundle row with hey move

[JoshAntBrown]

Fixes #388.

hey move takes box item IDs, and hey box view hands out two kinds under one ID space: threads, and bundle rows — a bundle stands in for one sender's whole stream in the box they are delivered to, and carries no topic_id. The move endpoint accepts either, so a bundle row moves like a thread, the command reports "1 thread moved", and the sender's next email arrives loose in the box the bundle just left. The reporter's batch of 31 IDs held 13 bundle rows; all 13 senders unbundled. HEY's own web app offers a bundle only Mark Seen, Note and Ignore, so the CLI was exposing a move the product withholds.

hey move now refuses a bundle row before moving anything. Each ID is checked against the bundles/unseen route — the discriminator hey bundle view and the thread-read fallback from #378 already rely on: it answers only for a bundle, a plain thread 404s, and a bundle read through still answers with its contact and no postings, so detection does not depend on unseen mail. A batch holding a bundle row is refused whole, naming every bundle ID, so the caller is never left guessing which of their IDs landed. The error points at hey contact bundle|unbundle for changing a sender's grouping; the long help, agent notes and README say why.

No new flag. Since the web app offers no way to move a bundle at all, an override would invent an escape hatch HEY itself does not provide. .surface is unchanged.

The cost is one extra GET per ID, bounded at 8 concurrent. That is the cheapest check the current API allows: the client exposes no posting lookup and no batch read, so the only way to learn a posting's kind from an ID is this route or paging the whole box. A server-side rejection in MovePostings would make the probe unnecessary and cover every client; if that is the direction, this can be dropped — the error shape, help text and tests would still apply to interpreting the server's refusal. Opened as a draft on that basis.

The check lives in bundle.go beside bundleNotFound, so one file owns the route. move_test.go's fake server now answers the bundles route, and its recorder is mutex-guarded for the concurrent probes; four tests cover the refusal, a mixed batch moving nothing, every bundle being named, and the help text. Verified against a live bundle before the test data was cleaned up: the refusal fires, a mixed batch moves nothing, normal threads are unaffected.

TMPDIR=/tmp/t make check green.

---

Summary by cubic

hey move now refuses bundle rows instead of treating them as threads, preventing a sender's stream from becoming unbundled when the row is moved. A batch containing any bundle row is rejected before the move request is sent.

Bug Fixes

• Detects bundle rows through the bundles/unseen endpoint and reports every offending ID with guidance to use hey contact bundle|unbundle.
• Adds one preflight request per ID, limited to eight concurrent requests, and fails closed if a check answers no page; regular thread moves continue as before.
• Updates command help, agent notes, README, and tests.

^{Written for commit 1e825db. Summary will update on new commits.}

[cubic-dev-ai]

All reported issues were addressed across 4 files

_{You’re at about 96% of the monthly reviewed-line limit. You may want to disable incremental reviews to conserve quota. Reviews will continue until that limit is exceeded. If you need help avoiding interruptions, please contact contact@cubic.dev.}

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[jeremy]

Looked this over against main for the maintainers' side of the draft question. The client-side probe is the right shape for now: bundles/unseen is the discriminator hey bundle view and the thread-read fallback already lean on, a bundle read through still answers, and one bounded GET per ID on a mutation that otherwise silently unbundles a sender is a proportionate cost. A server-side refusal in haystack's MovePostings would be welcome too, but it is a separate change in a separate repo and this does not have to wait for it — when it lands, the error shape and help here still apply.

Pushed 4b0db4a to keep the branch mergeable: rebased onto main, which moved the README's command reference to docs/cli.md — the bundle sentence now lives there — and folded in the cubic wording above. make check is green locally. The test workflow has not run on this PR yet (first-time-contributor approval), so the local run is the gate for now.

@JoshAntBrown, if this still matches what you intended, could you mark it ready for review? If we don't hear back a maintainer may carry it from here.

[JoshAntBrown]

Hey, thanks for the look and the rebase - that all matches what I intended.

[Copilot]

🟡 Changes recommended

The preflight currently fails open when the bundle endpoint returns a successful nil response.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Prevents hey move from relocating bundle rows and unintentionally unbundling future mail.

Changes:

• Adds concurrent bundle-row preflight checks.
• Rejects entire mixed batches and identifies offending IDs.
• Updates tests, command help, and CLI documentation.

[!TIP]
If you aren't ready for review, convert to a draft PR.
Click "Convert to draft" or run gh pr ready --undo.
Click "Ready for review" or run gh pr ready to reengage.

File summaries

File	Description
internal/cmd/move.go	Rejects detected bundles before moving.
internal/cmd/bundle.go	Implements bounded bundle detection and errors.
internal/cmd/move_test.go	Tests probes, rejection, and help text.
docs/cli.md	Documents bundle move refusal.

Review details

• Files reviewed: 4/4 changed files
• Comments generated: 1
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.