Jing Tao
Community@jingtao-lbl
Earth Research Scientist at LBNL.
Agent Skills by Jing Tao
Showing 47 vetted skills indexed across 1 GitHub repositories.
jingtao-lblphase3-diagnosis
Diagnose root causes of failing calibration targets in the A2MC workflow.
jingtao-lbladd-skill
Scaffolds and registers new skills across three registries with mechanical parity verification.
jingtao-lblecosim-run-workflow
Design, validate, submit, and score EcoSIM calibration ensembles across four parameter surfaces on HPC.
jingtao-lblcalibration-discipline
Enforces per-cycle and per-round completion checklists for long offline model calibration campaigns.
jingtao-lblsetup-discipline
Audits A2MC setup stages against executable definition-of-done checklists.
jingtao-lblwire-knowledge-graph
Audit and fix which curated seed relations reach a model's knowledge graph edges.
jingtao-lblpflotran-run-workflow
Design, submit, and score PFLOTRAN calibration ensembles against mass-balance output columns.
jingtao-lblarm-hpc-monitoring
Arms real-time log monitors on in-flight A2MC HPC ensemble processes at session start.
jingtao-lblphase2-screening
Rank calibration ensembles against validation targets and route results into Phase 3 diagnosis.
jingtao-lblonboard-model
Orchestrates onboarding a new environmental model into A2MC as a calibration-ready adapter.
jingtao-lblcalibration-goal
Drives the offline 7-phase calibration loop to convergence across turns and HPC waits.
jingtao-lbloffline-testing-workflow
Designs, launches, and analyzes FATES parameter-sweep experiments on HPC Morris base cases.
jingtao-lblwrite-report
Write comprehensive self-contained scientific reports for zero-context human readers.
jingtao-lblscientific-analysis
Runs open-ended calibration investigations that produce figures and evidence-cited analysis logs.
jingtao-lblrebuild-rag
Rebuilds and verifies per-model ChromaDB and NetworkX RAG indexes for A2MC.
jingtao-lblphase0-design
Designs, samples, and submits calibration ensembles for process-based environmental models.
jingtao-lblarm-local-monitoring
Monitors local no-scheduler A2MC ensemble runs via dispatch logs and process checks.
jingtao-lblinject-knowledge
Inject human-originated domain facts into A2MC's curated knowledge base channels.
jingtao-lblcompare-calibration-rounds
Compare A2MC calibration rounds via cross-round parameter, mechanism, and model-evolution ledgers.
jingtao-lblport-param-file
Migrate calibrated model parameter files across model API versions and formats.
jingtao-lblecosim-trait-check-refine
Validates and refines EcoSIM per-PFT plant trait values using deterministic sanity checks.
jingtao-lbladd-fates-parameter
Wire a new FATES parameter through EDParamsMod and the parameter file.
jingtao-lblbuild-rag-from-scratch
Constructs the RAG and GraphRAG knowledge layer from scratch for new or existing models.
jingtao-lblcherrypick-from-main
Cherry-pick generic commits from main into the adapter-kit branch with audit and verification gates.
Frequently Asked Questions About Jing Tao
FAQPage SchemaWhat tasks can I accomplish with Jing Tao's A2MC skills?▼
You can run end-to-end offline calibration of land-surface and reactive-transport models: design Morris/Sobol/LHS ensembles, screen against validation targets, root-cause failing targets, generate and test hypotheses, restart failed SLURM jobs, compare rounds, and produce cited reports and publication-grade figures.
Who are these skills designed for?▼
Earth-system research scientists and modelers calibrating FATES/ELM, EcoSIM, PFLOTRAN, or ATS against site observations. They assume familiarity with HPC schedulers, model parameter files, and sensitivity analysis, and support a human-in-the-loop workflow where the operator approves gates and curates knowledge.
How do I get started and run a calibration?▼
Run a2mc-init for first-run machine setup, then onboard-model for a new model or onboard-case for a new site. calibration-goal then drives the 7-phase loop to convergence, dispatching phase0 through phase6 skills, while arm-hpc-monitoring or arm-local-monitoring watches in-flight ensembles.
What infrastructure and dependencies are required?▼
Skills target Perlmutter-class SLURM HPC for CIME-submitted ELM/FATES ensembles, plus standalone-binary runs for EcoSIM, PFLOTRAN, and ATS. The knowledge layer uses ChromaDB vector stores and NetworkX graphs; document export needs pandoc with a LaTeX engine, and figures use matplotlib.
Are the skills model-specific or model-agnostic?▼
Most skills are model-agnostic with per-model backends: phase skills, calibration-goal, and round summaries resolve the producing tools per model, while dedicated skills cover model-specific traps such as EcoSIM version drift, PFLOTRAN deck scoring, and FATES EDParamsMod parameter additions.