GitHub issue #28058 in openai/codex, opened 2026-06-13 by ignatremizov, records a report that after #26210 (Encrypt multi-agent v2 message payloads, merged 2026-06-05, post-0.137.0), encrypted message delivery is the default for MultiAgentV2 spawn_agent, send_message, and followup_task, so users cannot inspect the exact delegated tasks, follow-up instructions, or intermediate child-to-parent messages as readable text in parent activity rows, child transcript/history, or after resuming a session; the report asks for a user-choosable readable delivery and a structured operational transcript with sender/recipient attribution, states the requested audit trail does not require exposing private reasoning, and notes that #35845 added conditional plaintext-response handling without changing the default, a position the body reaffirms in a source review dated 2026-10-26.
Verbatim Excerpts from Cited Records
“This issue remains unresolved in upstream's default MultiAgentV2 workflow.”
·
openai/codex (issue 28058 opening body, What issue are you seeing section)
·
2026-06-13T08:06:35Z
·GitHub Issue ↗
“Users still cannot inspect the exact tasks delegated by the parent, follow-up instructions, or intermediate child-to-parent messages when those payloads are encrypted.”
·
openai/codex (issue 28058 opening body, What issue are you seeing section)
·
2026-06-13T08:06:35Z
·GitHub Issue ↗
“The encrypted task/message payload is unavailable locally as readable text.”
·
openai/codex (issue 28058 opening body, What steps can reproduce the bug section)
·
2026-06-13T08:06:35Z
·GitHub Issue ↗
“Users should be able to choose readable inter-agent delivery and inspect a structured operational transcript with sender/recipient attribution.”
·
openai/codex (issue 28058 opening body, What is the expected behavior section)
·
2026-06-13T08:06:35Z
·GitHub Issue ↗
Counter-record: Issue open as of the fetch; 45 comments were read. Alek2077 (2026-07-04) reports the encrypted V2 path is forced for GPT-5.5 users on Codex 0.142.5, so the readable V1 delegation trail cannot be chosen through the documented feature setting. runfence (2026-07-14), embedding-shapes (2026-07-14, GPT 5.6 Sol Ultra), wuyak (2026-08-09, an output-side consequence in which a child evidence capsule was also unreadable), fwf-ben (2026-08-18, evals blocked without subagent prompts), jiahao-d (2026-09-01, Codex desktop for Windows, bundled codex-cli 0.151.0-alpha.7.2), future3OOO (2026-09-14, 0.154.0 on Linux/WSL2), saidelike (2026-10-29, App Server 0.164.0), and DanteMarshal (2026-11-13, 0.166.0 on Ubuntu 26.04.1 LTS) independently reproduce the missing readable task trail. Workarounds reported: not using subagents at all (Aquaticat, 2026-07-14) and forcing the V1 tools through a custom model catalog (ignatremizov, 2026-07-18). ignatremizov (2026-07-29) notes merged PR #35845 is a substantial partial fix confirming plaintext delivery is supported by the local orchestration path, while the body states the encrypted default remains; 92645417d9e5c763259dbebc306e3e (2026-07-29) notes the instruction type returned is always determined by the server side. No maintainer fix and no retraction of the report appear in the surveyed thread. The claim documents only the original report, not a verified current state.
Falsification probe: If GitHub issue #28058 is deleted without archive, if the author edits the opening body so the quoted sentences no longer match verbatim, or if the author posts a retraction of the report, this claim drops to Withdrawn.
Audit verification: Verified 100% continuous verbatim substring against official GitHub REST API by
scripts/verify_claim_verbatim.py.