| name | gumas-simulation-engine |
| description | Run, benchmark, and export the integrated GUMAS simulation engine (advanced default with legacy fallback) in Aurora workspaces. Use when users ask to run multi-turn simulations, compare seeds, evaluate stability/risk trends, export JSON state artifacts, or debug simulation behavior in GUMAS_SIM_2.5/SIM_ENGINE_OUTPUTS. |
| author | Aurora ORIONCORE |
GUMAS Simulation Engine
Quick Start
Run the bundled runner from the target workspace:
PYTHONPYCACHEPREFIX=/tmp \
python3 /Users/travisstreets/.codex/skills/gumas-simulation-engine/scripts/run_gumas_advanced.py \
--workspace "$PWD" --turns 20 --seed 42
Workflow
- Confirm
GUMAS_SIM_2.5/SIM_ENGINE_OUTPUTS exists in the chosen workspace.
- Run
scripts/run_gumas_advanced.py with requested turns and seed.
- Report the engine class used, last-turn summary, and output artifact path.
- If asked for comparison, rerun with alternate seeds and compare risk/stability trend deltas.
- If advanced import fails, use the legacy fallback notes in
references/runtime_notes.md.
Output Contract
Return:
- Engine mode and class (
GUMASAdvancedEngine or fallback)
- Turns requested/completed
- Last-turn summary string (or base tick counts in legacy mode)
- Output JSON path and whether export succeeded
Resources
- Runner script:
scripts/run_gumas_advanced.py
- Runtime notes:
references/runtime_notes.md
Workflow Position
- Upstream: direct simulation execution and comparison.
- Downstream:
aurora-narrative-tone-governor when prose artifacts need governance and aurora-exec-brief-pipeline when outputs must become leadership-ready.
- Preset reference:
/Users/travisstreets/.codex/skills/aurora-governance-orchestrator/references/workflow_presets.md.
