Codex review: needs maintainer review before merge. Reviewed June 21, 2026, 9:12 PM ET / 01:12 UTC.

Summary
The PR moves requester-session loading earlier in sendSubagentAnnounceDirectly, merges deliveryContextFromSession(requesterEntry) into the completion fallback origin, and adds same-channel plus cross-channel regression tests.

PR surface: Source +7, Tests +183. Total +190 across 2 files.

Reproducibility: yes. at source level: current main and v2026.6.9 compute deliverability before requester-session state can backfill to, and the linked issue shows a Telegram requester origin missing to. I did not rerun a live channel repro in this read-only review.

Review metrics: none identified.

Stored data model
Persistent data-model change detected: migration/backfill/repair: src/agents/subagent-announce-delivery.test.ts, serialized state: src/agents/subagent-announce-delivery.test.ts. Confirm migration or upgrade compatibility proof before merge.

Root-cause cluster
Relationship: fixed_by_candidate
Canonical: #86034
Summary: This PR is an open candidate fix for the missing-to requester-origin slice of the canonical generated-media completion delivery issue.

Members:

• canonical: Media generation succeeds but completion delivery fails and looks like generation failure #86034 - The PR uses closing syntax for this open issue and targets the Telegram missing-to mechanism documented in its comments.
• partial_overlap: Bug: False task failure reported when long-running media generation falls back to textual completion delivery #87906 - The closed issue covered the delivered-then-failed task-state contradiction, overlapping the media completion area but not this missing-target path.
• adjacent_distinct: Fix subagent DM completion delivery after yield #88182 - The merged PR changed direct-message subagent completion delivery after yield, but it fixed a different message-tool contract path.
• superseded: fix(agents): deliver DM media completions directly when requester handoff fails #95065 - The later overlapping PR was closed as a duplicate of this PR and its broader direct-DM fallback was not kept as the canonical path.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🐚 platinum hermit
Proof: 🦞 diamond lobster ✨ media proof bonus
Patch quality: 🐚 platinum hermit
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

• Get maintainer acceptance, bounding, or rejection of the same-channel requester-session fallback target contract.

Risk before merge

• [P1] Merging changes the failure mode from no external delivery to delivery through persisted same-channel requester-session route state; stale same-channel lastTo or thread state could target an older chat or thread unless maintainers accept that contract.
• [P1] The linked canonical issue also includes broader wake-false retry and structured media-handle/output-contract mechanisms, so this PR should close only the missing-to slice unless maintainers split or resolve the remaining work.

Maintainer options:

1. Accept Same-Channel Session Fallback
   A maintainer can accept requester-session lastTo and thread state as the recovery source when the completion origin lacks a target, relying on the existing cross-channel guard and tests.
2. Bound Stale Route State
   If persisted same-channel state is too risky, require a freshness or route-binding check before using session state as the generated-media fallback target.
3. Keep The Contract Open
   If maintainers reject requester-session fallback entirely, pause or close this PR and keep the delivery-target contract decision on the canonical issue.

Next step before merge

• [P2] The remaining action is maintainer judgment on the fallback delivery-target contract, not a narrow automated code repair.

Security
Cleared: The diff changes TypeScript delivery-context selection and focused tests only; no dependency, workflow, secret, package, or supply-chain surface is changed.

Mantis Telegram Desktop Proof

Summary: Mantis captured native Telegram Desktop before/after GIF evidence: main does not show the generated image completion, while this PR delivers it back to the chat.

• Scenario: telegram-desktop-proof
• Trigger: workflow_dispatch
• Run: https://github.com/openclaw/openclaw/actions/runs/27971069903
• Artifact: https://github.com/openclaw/openclaw/actions/runs/27971069903/artifacts/7800625529
• Baseline: pass at f00f42abf7abbafe534d90e0b30e761314b6e0fc, expected baseline visual proof captured
• Candidate: pass at 0010944da9ba7fdf8452f8cd24d1cb35c8342187, expected candidate visual proof captured
• Overall: true

Main screenshot	This PR screenshot

Main	This PR

Motion-trimmed clips:

• Main MP4
• This PR MP4

Raw QA files: https://artifacts.openclaw.ai/mantis/telegram-desktop/pr-89949/run-27971069903-2/index.json

CI feedback request

check-lint — need help locating the errors

The lint job run 26902096731 reports:

##[error]Unnecessary escape character '/'
##[error]Unnecessary escape character '/'
##[error]Unnecessary escape character '/'
##[error]Unnecessary escape character '/'

Found 0 warnings and 4 errors.
Finished in 14.7s on 9563 files with 242 rules using 8 threads.

The CI log surfaces the rule but no file:line locations. I verified locally:

• Running the same script (node scripts/run-oxlint-shards.mjs --threads=4) on oxlint 1.67.0 against this branch produces 0 errors, 0 warnings.
• Running the same script against the PR's merge-base (286e5ffe) also produces 0 errors, 0 warnings.
• grep -nE '\\/' over the two files this PR touches (src/agents/subagent-announce-delivery.ts, src/agents/subagent-announce-delivery.test.ts) returns no matches.

This suggests one of:

1. The CI runner is on a different oxlint version where no-useless-escape finds something local 1.67.0 does not (most likely — the action prints with 242 rules; my local shows 242 rules too, but the CI's 9563 files vs my local count may differ).
2. The 4 errors live in a path my local lint isn't scanning (some shard or workspace I'm not iterating).

Could a maintainer either:

• re-run the lint with --silent=false / annotation mode disabled so file paths surface, or
• confirm the oxlint binary version + invocation used in CI so I can reproduce.

I'm happy to fix the escape if it's something this PR introduced; I just can't see where to fix without the location.

Real behavior proof — addressed

Mantis run 26902454818 captured a Telegram Desktop end-to-end with Overall: true (baseline + candidate both pass). I've updated the PR body to reference it under "Real behavior proof", so the policy check should pass on re-run.

@clawsweeper re-review

PR body updated with live Mantis Telegram Desktop proof (run 26902454818, Overall: true, both baseline + candidate pass). The earlier 'Real behavior proof' policy check ran before Mantis published, so the artifact link was not yet in the body — that should now satisfy the proof gate.

🦞🧹
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: Complete
• Detail: The targeted re-review finished, the durable review comment was updated, and the synced verdict was routed.
• Run: https://github.com/openclaw/clawsweeper/actions/runs/26924320918
• Updated: 2026-06-04T01:39:23.190Z

@clawsweeper re-review

@clawsweeper re-review

Land-ready verification for 0010944da9ba7fdf8452f8cd24d1cb35c8342187:

• Fresh autoreview: clean; no accepted/actionable findings after route-account, stale-thread, and internal-origin fixes.
• Media background family plus subagent announce delivery: 138 tests passed across five files.
• Core test typecheck passed via node scripts/run-tsgo.mjs -p test/tsconfig/tsconfig.core.test.json --incremental --tsBuildInfoFile /tmp/pr89949-core-test.tsbuildinfo.
• Exact-head GitHub CI: all 63 relevant checks green in https://github.com/openclaw/openclaw/actions/runs/27969100551.
• Exact-head native Telegram Desktop proof: https://github.com/openclaw/openclaw/actions/runs/27971069903 attempt 2 succeeded. Its direct production-runtime probe shows main omitting the generated image completion and candidate delivering it to the requester chat; paired screenshots/GIFs and raw evidence are published at fix(media): pin requester delivery route when task starts #89949 (comment).

Known scope: this lands only the missing-target mechanism. Retry asymmetry and other #86034 mechanisms remain open.

Landed via rebase onto main.

• Scoped tests: node scripts/run-vitest.mjs src/agents/tools/media-generate-background-shared.test.ts src/agents/tools/image-generate-background.test.ts src/agents/tools/video-generate-background.test.ts src/agents/tools/music-generate-background.test.ts src/agents/subagent-announce-delivery.test.ts (5 files / 138 tests passed)
• Changelog: skipped; release generation owns CHANGELOG.md for normal PRs
• Land commit: 6a2b993
• Merge commit: 780f83b

Thanks @wangwllu!