Testing note for #83365:

On this Lobster machine, the reproduced failure was specific to group media delivery through the existing rich/file path:

• OpenClaw generated a valid PNG for the group request.
• The normal final media reply did not appear as media in the group.
• A direct imsg send-rich --chat <redacted-group> --file <generated-png> attempt failed locally with an attachment-directory permission error.
• A direct imsg send-attachment --chat <redacted-group> --file <generated-png> --transport auto attempt succeeded and produced an outbound attachment in the group.

This PR therefore changes OpenClaw's explicit-chat, media-only iMessage send path to use send-attachment --transport auto, matching the command path that succeeded during the live repro. This does not claim that every send-rich --file scenario is broken; it documents that send-rich --file failed for the tested iMessage group on this machine while send-attachment --transport auto worked.

ClawSweeper status: review started.

I am starting a fresh review of this pull request: Fix iMessage group media delivery This is item 1/1 in the current shard. Shard 0/1.

This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking.

Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted.

Live validation after this PR was built locally from commit 6e10d32 and the gateway was restarted from /Users/lobster/GitHub/openclaw/dist/index.js:

• Controlled media-only OpenClaw send to the new redacted iMessage test group used the existing generated PNG at ~/.openclaw/media/tool-image-generation/red-coffee-mug-blue-table---a46a540c-2275-4db2-bdd2-e6bccd361739.png.
• Messages DB evidence: a new outbound row was created in the group chat with attachment_count=1.
• Controlled media-only OpenClaw send to the redacted 1:1 DM using the same PNG also created a new outbound row in the DM chat with attachment_count=1.
• The generated file was optimized for delivery during both sends (Optimized media from 1.14MB to 0.05MB).

This validates the PR's media-only iMessage delivery path for both an explicit group chat_id target and an explicit DM chat_id target on this machine. Identifiers, handles, phone numbers, and private chat names are intentionally omitted.

@clawsweeper re-review

Fresh proof is now on the PR: Lobster built commit 6e10d32, restarted the gateway from the built dist, and validated media-only iMessage sends for both an explicit group chat_id target and an explicit DM chat_id target with outbound DB rows showing attachment_count=1. Please review against the updated proof and non-draft state.

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Re-review progress:

• State: Failed
• Detail: The targeted re-review did not finish cleanly. Check the workflow run for details.
• Run: https://github.com/openclaw/clawsweeper/actions/runs/26428452532
• Updated: 2026-05-26T02:18:20.800Z

Maintainer verification before merge:

• Local cleanup check on follow-up commit c92d111543: pnpm run lint:extensions:bundled in the Fix iMessage group media delivery #86720 worktree, passed.
• GitHub checks on c92d111543: gh pr checks 86720 --json name,state,link --jq '[.[] | select(.state != "SUCCESS" and .state != "SKIPPED" and .state != "NEUTRAL") | {name,state,link}]' returned [].
• Relevant CI/proof runs observed green include CI run 26428503154, Real behavior proof run 26428501870, OpenGrep PR Diff run 26428503150, and CodeQL Critical Quality run 26428503100.
• Live behavior proof: Lobster built the PR, restarted the gateway from built dist, and verified media-only iMessage sends to an explicit group chat_id and explicit DM chat_id, both producing outbound DB rows with attachment_count=1.
• Known proof gap: the live behavior proof was on original behavior commit 6e10d32; the follow-up c92d111543 only adds braces around existing timer cleanup to satisfy lint and does not change send behavior.

No changelog entry added per maintainer direction.