muse-runner

Muse Runner

Execute prompts as the Muse seat through the shared cline-runner implementation. The next consumer receives the shared runner envelope. The run is done when the envelope confirms the Muse model or reports the seat unavailable.

Default Model

meta/muse-spark-1.1 is the single default model declaration for this skill. OpenRouter lists Muse Spark 1.1 with a 1M token context window and United States only availability.

Prerequisites

1. Install cline in PATH (npm install -g cline).
2. Authenticate a Cline provider through cline auth that can resolve the default model.
3. Use an OpenRouter account and request location that meet the model's United States availability rule.

Security Model

This skill uses the Cline execution and data sharing contract. Prompt text, prompt files, metadata, and files read during the run may be sent to the selected provider.

Passing --model changes the selected provider's persisted default in ~/.cline/data/settings/providers.json. Use --data-dir when the run must not change that default. A new data directory needs its own authenticated provider.

Analysis roles use native --auto-approve false by default. Pass --allow-write only when the role must change files.

Usage

SKILL_DIR="<absolute path of the directory containing this SKILL.md>";
python3 "$SKILL_DIR/scripts/run_muse.py" "your prompt here"

Examples

SKILL_DIR="<absolute path of the directory containing this SKILL.md>";
python3 "$SKILL_DIR/scripts/run_muse.py" "Explain this module" --role researcher

SKILL_DIR="<absolute path of the directory containing this SKILL.md>";
python3 "$SKILL_DIR/scripts/run_muse.py" --prompt-file /tmp/review.md --role codereviewer --json

Behavior

1. Delegate to cline-runner with runner=muse and effective_runner=cline.
2. Forward the default model through native --model on every run unless the caller overrides it.
3. Use Cline NDJSON output and the shared result envelope.
4. Preserve seat fidelity. The Muse seat returns a Muse receipt or reports an observable failure. It never substitutes another model.

State the result as: "The Muse seat preserved seat fidelity and returned <effective_model>," or name the blocking error.

Acceptance Contract

The runner is ready when:

1. --help exits with code 0.
2. A missing Cline binary returns runner=muse, effective_runner=cline, return_code=-2, and status=seat_unavailable.
3. A successful live run returns native_model_id=meta/muse-spark-1.1.