swmm-experiment-auditlisted

# SWMM Experiment Audit ## What this skill provides - A standard audit layer for Agentic SWMM runs. - Consolidation of dispersed `manifest.json`, QA JSON, logs, metrics, and artifact paths. - Machine-readable outputs for reproducibility and review. - Obsidian-compatible Markdown notes for human research records. - Default local Obsidian export into a clean English audit vault. - Automatic update of the Obsidian `Experiment Audit Index`. - Optional run-to-run comparison for baseline/scenario or before/after parser validation. This skill records what happened. It does not run SWMM, build models, invent missing artifacts, or replace module-level validation. ## When to use this skill Use this skill after any of these events: - `swmm-end-to-end` completes successfully. - `swmm-end-to-end` stops or fails after producing partial artifacts. - A user wants an Obsidian-ready experiment note for an existing run directory. - A user wants to compare two run directories. - A run needs evidence for reproducibility, metric provenance, QA status, or paper claims. Do not use this skill as a substitute for `swmm-runner`, `swmm-builder`, or calibration tools. Run the model first, then audit the run directory. ## Output contract For every audited run, write these files into the run directory unless explicit output paths are provided: - `experiment_provenance.json` - `comparison.json` - `experiment_note.md` `experiment_provenance.json` is the machine-readable source for: - run identity