gpt-cmdr

Create deterministic paired RASMapper and RAS Commander MapLibre evidence bundles from one review specification. Use when comparing a HEC-RAS model in RASMapper with its web viewer, validating map/tree/legend fidelity on desktop and mobile, checking nodata/extent/default layers, or establishing visual regression fixtures with fixed numeric COG probes.

Capture HEC-RAS and standalone RASMapper windows through current ras-commander Win32 APIs for documentation, debugging, QA/QC evidence, galleries, and deterministic spatial review packages. Use for screenshot, visible-window capture, RASMapper snapshot, dialog capture, or model gallery requests on Windows.

Use RASMapper layer and view automation for spatial QA/QC, contextual model review, and documentation screenshots. Launches standalone RasMapper.exe, toggles reference maps, basemaps, terrain, results, and geometry elements, zooms to HDF-derived extents, enables Update Legend with View, and captures snapshots for agent review.

Pause agent work and request explicit human input at hard gates where professional judgment, public API compatibility, HEC-RAS execution, agent infrastructure, or git/release operations create material risk. Triggers: human-in-the-loop, HITL, human approval, human decision, approval gate, hard gate, public API change, engineering assumption, HEC-RAS execution approval, agent infrastructure change, release gate.

Codex-native adapter for invoking Anthropic Claude Code CLI as an independent read-only QAQC reviewer using markdown file handoff. Use when the user explicitly asks Codex to ask Claude, Claude Code, Opus, or Anthropic for a QAQC/code review, second-model review, or independent review. Do not trigger for generic QAQC unless the user asks for Claude specifically. Triggers: claude review, claude code review, ask claude, invoke claude, claude qaqc, opus review, anthropic review, second model with claude Prerequisites: Claude Code CLI authenticated and available as `claude`. Default model: opus with xhigh effort for deep reviews; use sonnet/high for faster or lower-cost review passes.

Enforce the feature branch workflow — intercept any attempt to merge, push, or commit directly to main and redirect to the correct feature branch process. Triggers on "merge to main", "push to main", "PR to main", or any operation targeting the main branch.

Delegate code tasks to OpenAI Codex CLI using markdown file handoff pattern. Write instructions to TASK.md, Codex reads and executes, outputs to OUTPUT.md. Use for implementation, refactoring, code generation requiring extended thinking. Triggers: codex, codex cli, delegate to codex, codex subagent, implementation task, refactoring task, code generation, gpt-5.2-codex, openai codex, extended thinking, multi-file changes, complex implementation Prerequisites: Codex CLI authenticated (codex login or OPENAI_API_KEY) Model: gpt-5.2-codex (default, supports xhigh reasoning effort)

Legacy Claude-only provider orchestration for Google Gemini CLI using markdown file handoff. Write review request to REVIEW.md, Gemini analyzes, outputs findings to FINDINGS.md. Use only when the user explicitly requests Gemini. Do not trigger for generic QAQC, quality assurance, security review, architecture review, or second-opinion requests. Triggers: gemini, gemini cli, delegate to gemini, gemini subagent, gemini-3-pro-preview, gemini-3-flash-preview Prerequisites: Gemini CLI authenticated (gemini login or GEMINI_API_KEY) Models: gemini-3-pro-preview (default), gemini-3-flash-preview (large context)

Codex-native adapter for invoking Anthropic Claude Code CLI as an independent read-only QAQC reviewer using markdown file handoff. Use when the user explicitly asks Codex to ask Claude, Claude Code, Opus, or Anthropic for a QAQC/code review, second-model review, or independent review. Do not trigger for generic QAQC unless the user asks for Claude specifically. Triggers: claude review, claude code review, ask claude, invoke claude, claude qaqc, opus review, anthropic review, second model with claude Prerequisites: Claude Code CLI authenticated and available as `claude`. Default model: opus with xhigh effort for deep reviews; use sonnet/high for faster or lower-cost review passes.

Claude-native adapter for invoking OpenAI Codex CLI through markdown handoff files. Use when Claude Code needs an explicit Codex QAQC review, Codex code review, cross-harness verification, implementation pass, refactoring pass, or second-model opinion. Default to read-only review unless the user explicitly asks Codex to edit files. Uses current `codex exec` commands, repository-local paths, and TASK.md/OUTPUT.md artifacts under `agent_tasks/handoffs/codex/`.

Clones HEC-HMS components (basins, met models, control specs, runs) using the CLB Engineering LLM Forward approach. Creates non-destructive, traceable, GUI-verifiable copies for QAQC comparison, scenario analysis, and parameter sensitivity testing. Use when creating alternative scenarios, setting up QAQC workflows, comparing baseline vs updated models, preserving originals while testing modifications, Atlas 14 precipitation updates (keep TP40 baseline), or model calibration with different parameter sets. All clones appear in HEC-HMS GUI with separate DSS outputs for side-by-side comparison. Trigger keywords: clone, duplicate, copy, QAQC, scenario, alternative, baseline, comparison, side-by-side, non-destructive, traceable.

Executes HEC-HMS simulations using HmsCmdr.compute_run(), handles parallel execution across multiple runs, and manages Jython script generation. Use when running HMS simulations, executing runs, computing models, setting up parallel computation workflows, generating Jython scripts, or running batch simulations. Handles HMS 3.x and 4.x version differences and Python 2/3 compatibility. Trigger keywords: run simulation, execute HMS, compute run, parallel execution, batch runs, Jython script, HMS compute, run model.

Exports HEC-HMS basin geometry and simulation results to cloud-native geospatial formats (GeoParquet, PMTiles) using the hms2cng package (HMS to Cloud Native GIS). Supports all HMS element types (subbasins, reaches, junctions, diversions, reservoirs, sources, sinks, watershed boundary) plus SQLite-based grid layers (subbasin polygons, flowpaths). Also exports results summary statistics (peak, min, mean) spatially joined with geometry. Use when generating GeoParquet for DuckDB/web analysis, PMTiles for map visualization, syncing to PostGIS, exporting basin elements for web GIS, building interactive map layers, or creating cloud-optimized archives of HMS project geometry. Trigger keywords: GeoParquet, PMTiles, cloud native, vector tiles, export geometry, export results, spatial export, parquet, tippecanoe, PostGIS sync, web map, DuckDB query, hms2cng.

Extracts and analyzes HEC-HMS simulation results from DSS files using HmsDss and HmsResults classes. Handles peak flows, hydrographs, volume summaries, and time series data. Leverages ras-commander's RasDss for DSS V6/V7 support. Use when processing HMS results, extracting peak flows, analyzing hydrographs, computing volumes, exporting time series, comparing multiple runs, or linking HMS results to HEC-RAS boundary conditions. Trigger keywords: DSS file, results, peak flow, hydrograph, time series, volume, extract results, HMS output, analyze results, compare runs.

Investigates HEC-HMS internals through decompiled Java classes. Provides JythonHms API reference, CLI options, version-specific differences (3.x vs 4.x), and guides on-demand decompilation for new discoveries. Use when debugging HMS behavior, discovering undocumented features, understanding version differences, validating automation approaches, checking JythonHms method signatures, or investigating HMS API capabilities. Complements hms_query_docs (official docs) with internal implementation details. Do NOT use for: official documentation queries (use hms_query_docs), hms-commander API questions (read code docstrings), or project file parsing (use hms_parse_basin-models). Trigger keywords: decompile, HMS internals, JythonHms API, HMS version differences, CLI options, undocumented, HMS jar, debugging HMS, HMS source code, method signatures.

Links HEC-HMS watershed models to HEC-RAS river models by extracting HMS DSS results and preparing them for RAS boundary condition import. Handles flow hydrograph export, spatial referencing (HMS outlets to RAS cross sections), DSS pathname formatting, quality validation, and time series alignment. Use when setting up HMS-to-RAS workflows, exporting HMS results for RAS, preparing upstream boundary conditions, coordinating watershed-to-river integrated modeling, or validating HMS results before handoff. Leverages shared RasDss infrastructure for consistent DSS operations across both tools. Trigger keywords: HMS to RAS, link HMS RAS, boundary condition, upstream BC, watershed to river, integrated model, export HMS, spatial matching, hydrograph import.