gemini-cli-runtime

Gemini Runtime

Use this skill only inside the gemini:gemini-rescue subagent.

Primary helper:

• node "${CLAUDE_PLUGIN_ROOT}/scripts/gemini-companion.mjs" task "<raw arguments>"

Execution rules:

• The rescue subagent is a forwarder, not an orchestrator. Its only job is to invoke task once and return that stdout unchanged.
• Prefer the helper over hand-rolled CLI strings, git operations, or any other Bash activity.
• Do not call adversarial-review, status, result, or cancel from gemini:gemini-rescue.
• Use task for every rescue request, including diagnosis, planning, research, and explicit fix requests.
• You may use the gemini-prompting skill to rewrite the user's request into a tighter prompt before the single task call.
• That prompt drafting is the only Claude-side work allowed. Do not inspect the repo, solve the task yourself, or add independent analysis outside the forwarded prompt text.
• Leave --effort unset unless the user explicitly requests a specific effort.
• Leave model unset by default. Add --model only when the user explicitly asks for one.
• Model aliases (source of truth: MODEL_ALIAS_ENTRIES in plugins/gemini/scripts/lib/model-map.mjs): flash / flash3 → gemini-3-flash-preview, pro / pro3 → gemini-3.1-pro-preview, lite3 → gemini-3.1-flash-lite, flash25 → gemini-2.5-flash, pro25 → gemini-2.5-pro, lite / fast → gemini-2.5-flash-lite.
• Read-only is the default; add --write only when the user explicitly asked for edits.

Command selection:

• Use exactly one task invocation per rescue handoff.
• If the forwarded request includes --background or --wait, treat that as Claude-side execution control only. Strip it before calling task.
• If the forwarded request includes --model, normalize aliases and pass it through.
• If the forwarded request includes --effort, pass it through to task.
• If the forwarded request includes --timeout <seconds>, pass it through to task. Whole seconds, 30 to 3600. It bounds both the run's duration and the amount of output it can emit before being killed.
• --effort accepted values: Gemini engine accepts none, minimal, low, medium, high, xhigh; AGY engine accepts only low, medium, high.
• --resume: add --resume-last, even if the request text is ambiguous.
• --fresh: use a fresh task run, do not add --resume-last.
• task --resume-last: for "keep going", "resume", "apply the top fix", or "dig deeper".

Engine routing:

• Gemini CLI and AGY are both first-class supported engines; only the selected engine's CLI is required.
• Default engine: credential-gated auto-detection (selects Gemini if installed and authenticated with valid credentials, otherwise AGY).
• Force AGY: add --engine agy.
• Force Gemini CLI: add --engine gemini.
• Note for AGY engine: --model takes an exact model ID from agy models or --effort accepts low, medium, or high; the two cannot be combined, and Gemini model aliases are rejected for AGY.

Safety rules:

• Read-only is the default in gemini:gemini-rescue; add --write only when the user explicitly asked for edits.
• Preserve the user's task text as-is apart from stripping routing flags.
• Do not inspect the repository, read files, grep, monitor progress, poll status, fetch results, cancel jobs, summarize output, or do any follow-up work of your own.
• Return the stdout of the task command exactly as-is.
• If the Bash call fails or the engine cannot be invoked, return nothing.