docs(releases): v1.0.0 code-freeze plan

[vybe]

Output of /release-plan (full mode) for the 1.0 cut, run on rc3 (15e3502c0). It is the work order for the 1.0 freeze, so /release can diff intent against actual at cut time.

Go/No-Go on rc3: NO-GO. GO on rc4 once MUST LAND clears:

• M1: merge fix(skill-gate): the person who asked sees how a gated request ended (#3653) #3657 / fix: the Chat tab notice of a finished gated run quotes the run's reply (#3654) #3658 (t#3653, t#3654). These are the two A2A-row P1s ruled fix before cut.
• M2: make the keyless trusted-network A2A path inert. The operator ruled on ent#862 that it must not ship in 1.0.
• M3/M4: tag rc4, then re-test A2, B3, B7, B11 and B14.
• M5: operator rulings R1–R3.
• M6: write the 1.0 release notes, including 7 default-ON upgrade behaviour changes found by the dark-flag scan.

The seam scans found the migration tracks (0064–0098), the agent-server mirrors and the MCP reach all clean, apart from the A2A trusted-networks fracture above.

Docs only; no code change.

🤖 Generated with Claude Code

[trinity-ability]

merge-train: batch validated on train/20261013-2303 (#3685)

[AndriiPasternak31]

@vybe, for the release notes (M6): the draft blocks are missing one required upgrade step.

#3304 (credential files are owner-tier to read; in rc3 and rc4) carries a required release step: after deploying, stop and start every agent through Trinity, so each one is recreated on the new base image. Until an agent is recreated, a download or preview of its files by anyone below its owner or an admin answers 403 agent_restart_required. That includes people the agent is shared with and agent keys, so it also covers the MCP tools list_agent_pipelines and get_agent_pipeline_state, which orchestrator agents use to read sibling pipelines.

The draft Known-limitations line "Recreate agents on the 1.0 base image" says older-image agents "keep working" and names only two protections, and the Upgrade-notes block doesn't mention the step at all.

Proposed line for the Upgrade-notes block (public-safe):

- **Stop and start every agent after upgrading.** This recreates each agent on the 1.0
  base image. Until an agent is recreated, people it is shared with and agent keys
  (including the `list_agent_pipelines` and `get_agent_pipeline_state` tools) cannot
  download or preview its files: they get 403 `agent_restart_required`. Its owner and
  admins are not affected.

The Known-limitations line could then read "Agents still on an older image keep working for their owner and admins, but…" so the two blocks agree.