rescue

Invoke the gemini:gemini-rescue subagent via the Agent tool (subagent_type: "gemini:gemini-rescue"), forwarding the raw user request as the prompt.

gemini:gemini-rescue is a subagent, not a skill — do not call Skill(gemini:rescue) (no such skill) or Skill(gemini-rescue) (that re-enters this command and hangs the session). The command runs inline so the Agent tool stays in scope; forked general-purpose subagents do not expose it.

Raw user request: $ARGUMENTS

Execution mode:

• If the request includes --background, run the gemini:gemini-rescue subagent in the background.
• If the request includes --wait, run the gemini:gemini-rescue subagent in the foreground.
• If neither flag is present, default to foreground.
• --background and --wait are execution flags for Claude Code. Do not forward them to task, and do not treat them as part of the natural-language task text.
• --model, --effort, and --plan are runtime-selection flags. Preserve them for the forwarded task call, but do not treat them as part of the natural-language task text.
• If the request includes --resume, do not ask whether to continue. The user already chose.
• If the request includes --fresh, do not ask whether to continue. The user already chose.
• Otherwise, before starting Gemini, check for a resumable rescue thread from this Claude session by running:

node "${CLAUDE_PLUGIN_ROOT}/scripts/gemini-companion.mjs" task-resume-candidate --json

• If that helper reports available: true, use AskUserQuestion exactly once to ask whether to continue the current Gemini thread or start a new one.
• The two choices must be:
  • Continue current Gemini thread
  • Start a new Gemini thread
• If the user is clearly giving a follow-up instruction such as "continue", "keep going", "resume", "apply the top fix", or "dig deeper", put Continue current Gemini thread (Recommended) first.
• Otherwise put Start a new Gemini thread (Recommended) first.
• If the user chooses continue, add --resume before routing to the subagent.
• If the user chooses a new thread, add --fresh before routing to the subagent.
• If the helper reports available: false, do not ask. Route normally.

Operating rules:

• The subagent is a thin forwarder only. It should use one Bash call to invoke node "${CLAUDE_PLUGIN_ROOT}/scripts/gemini-companion.mjs" task ... and return that command's stdout as-is.
• Return the Gemini companion stdout verbatim to the user.
• Do not paraphrase, summarize, rewrite, or add commentary before or after it.
• Do not ask the subagent to inspect files, monitor progress, poll /gemini:status, fetch /gemini:result, call /gemini:cancel, summarize output, or do follow-up work of its own.
• Leave --effort unset unless the user explicitly asks for a specific reasoning effort. Note: the flag is accepted but has no effect yet (pending upstream ACP support for thinkingLevel).
• Leave the model unset unless the user explicitly asks for one. Accept Gemini model aliases: pro, pro-3, flash, flash-lite, 2.5-pro, auto. Pass concrete model IDs through unchanged.
• --plan maps to Gemini's --approval-mode plan (read-only planning run). Without --plan, the companion defaults to --yolo --sandbox (Gemini CLI built-in sandbox, filesystem scoped to the working tree).
• Leave --resume and --fresh in the forwarded request. The subagent handles that routing when it builds the task command.
• If the helper reports that Gemini is missing or unauthenticated, stop and tell the user to run /gemini:setup.
• If the user did not supply a request, ask what Gemini should investigate or fix.
• When the rescue is exploratory or diagnostic (e.g., "where is X used?", "trace this flow", "why is this failing?"), the forwarded prompt should suggest that Gemini reach for its built-in codebase_investigator subagent before drafting a fix — it ships with the codebase-investigation skill and is already configured to use gemini-3.1-pro-preview.