feat(ux): show resolved model/effort in the /codex:rescue subagent launch line #34

Problem

When Claude Code launches the codex:codex-rescue subagent (foreground or background), the launch line shows only the topic:

codex:codex-rescue(Codex консультация по #144)

The model and effort Codex actually uses are nowhere visible to the user — especially painful in background runs, where the user sees only "Codex started in the background" and can't tell which model/effort was picked. The companion already has announceRun(model, effort) (PR #24) writing Codex: model=X effort=Y to stderr, but stderr is swallowed in background runs.

Where the launch line comes from

The parentheses are the description parameter of the Agent({subagent_type, description, prompt }) tool call. description is free-form text that Claude Code (the caller) composes when launching the subagent — before the subagent selects the model. Today commands/rescue.md instructs the caller to write a short topic, with no rule about model/effort, so none appears.

Because codex-rescue.md (lines 31-32) selects the model only from explicit user flags (--model/--effort) or leaves it unset (default), the caller can resolve the same value deterministically before launch — no desync risk. (No inherited/hidden model logic to reproduce.)

Proposed fix

Add an instruction to the rescue command/skill so the caller appends the resolved model and effort to the Agent-tool description:

codex:codex-rescue(Codex консультация по #144 model=gpt-5.6-sol effort=xhigh)

• Resolve aliases to full ids (the way the companion does): spark → gpt-5.3-codex-spark, sol → gpt-5.6-sol, terra → gpt-5.6-terra, luna → gpt-5.6-luna. Concrete ids (e.g. gpt-5.4-mini) pass through.
• When the user specified nothing: model=default effort=default.
• Companion is not changed — announceRun stays for foreground-stderr; the launch line is covered by the skill layer.

Files

• plugins/codex/commands/rescue.md — add the description-formatting rule
• plugins/codex/skills/codex-cli-runtime/SKILL.md — same rule for direct skill invocations
• Possibly plugins/codex/agents/codex-rescue.md — if description is shaped elsewhere

Verification

Manual: /codex:rescue --model sol --effort xhigh <prompt> → launch line shows model=gpt-5.6-sol effort=xhigh. /codex:rescue <prompt> (no flags) → model=default effort=default.