Jeden Skill in Manus ausführen
mit einem Klick

Jeden Skill in Manus mit einem Klick ausführen

phase0-design

Sterne6

Forks1

Aktualisiert7. Juli 2026 um 07:52

Run Phase 0 (DESIGN) of the A2MC calibration workflow as the offline agent — the human-in-the-loop analog of the orchestrator's `_run_design()`. Sample the parameter space (Morris/Sobol/LHS), materialize one FATES parameter file per case, generate + build + submit the ensemble, and arm monitoring. Use when the user says "design a new round", "set up the ensemble", "sample the parameters", "submit the Morris ensemble", "start a calibration round", "expand the parameter space (redesign)".

Installation

Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.

In Manus ausführen

Quelle

jingtao-lbl

jingtao-lbl/A2MC-elm

GitHub-Repository öffnen Creator-Repositorys ansehen

Download

In Manus ausführen

Verwandte BerufeSOC

Basierend auf der SOC-Berufsklassifikation

SoftwareentwicklerInformatik- und Mathematikberufe·SOC 15-1252

SKILL.md

readonly

name	phase0-design
description	Run Phase 0 (DESIGN) of the A2MC calibration workflow as the offline agent — the human-in-the-loop analog of the orchestrator's `_run_design()`. Sample the parameter space (Morris/Sobol/LHS), materialize one FATES parameter file per case, generate + build + submit the ensemble, and arm monitoring. Use when the user says "design a new round", "set up the ensemble", "sample the parameters", "submit the Morris ensemble", "start a calibration round", "expand the parameter space (redesign)".
modes	{"requires_fates":false,"nutrient_pathway":"any","scope":["calibration"],"summary":"Offline analog of online Phase 0 (design/sample/submit the ensemble). Applies in every calibration mode."}

Phase 0: Design & Submit (offline agent)

The offline analog of the orchestrator's _run_design(). Online this is deterministic sampling + submission; offline you own the design decisions the online loop takes from config — scheme, trajectory count, which parameters vary, bounds, and whether this is a fresh round or a redesign/replay. The underlying scripts are the same the online agent drives.

Floor, not ceiling. This skill is the minimum to stand up a round the way the online agent would. As the offline agent you can do more: reconsider the parameter list itself, add or widen bounds a redesign implies, sanity-check the sampling against prior-round μ* before spending compute, or design a targeted sub-ensemble the fixed loop would never propose. Match the discipline (verify param existence, don't clobber the committed matrix, validate before submit), then exercise judgment about what to sample.

Deliverable: simulation jobs queued (ADSP→RGSP→TRANS dependency chains), submission_manifest.json written, monitoring armed.

Step 1 — sample the parameter space (parameter-file prep — no simulations)

phases/phase0_design/create_parameter_sample.py --method {morris|sobol|lhs} reads $A2MC_PARAM_LIST_FILE (name + bounds) and writes the ensemble matrix ($A2MC_ENSEMBLE_MATRIX_FILE) and SALib problem ($A2MC_SALIB_PROBLEM_FILE). Morris: --trajectories T (N = T×(P+1)).

Footgun — matrix overwrite. $A2MC_ENSEMBLE_MATRIX_FILE for Kougarok points at the committed, byte-stable 4890-set matrix the manuscript depends on. To regenerate without clobbering it, write to a scratch path (--output-matrix) or repoint the env var to a fresh round file first.

Redesign / replay (reuse a prior round's NC dir instead of fresh sampling): apply_param_override.py (global override, e.g. R5's prescribed_puptake=1.0 onto R3's NCs) or create_subset_replay.py (top-N replay preserving source case numbers).

Step 2 — materialize per-case parameter files (parameter-file prep — no simulations)

phases/phase0_design/generate_parameter_files.py reads the X matrix + $A2MC_BASE_PARAM_FILE and writes one parameter file per row into $A2MC_PARAM_DIR (pattern $A2MC_PARAM_PATTERN, {N} placeholder). It is method- and format-agnostic (JSON at api-43+, NetCDF at api-31 and earlier) and dispatches to tools/modify_fates_parameters.py for the per-parameter edits. Verify file count = N_sets.

Footgun — column↔parameter mapping + file format. The X-matrix column order must match the site parameter list; build_param_lookup() in tools/modify_fates_parameters.py maps each shorthand → FATES name + PFT + organ. Confirm tools/describe_mode.py shows the milestone you expect so the right parameter-file format (JSON vs NetCDF) is written. Never assume a column's meaning.

Step 3 — generate scripts, build, submit the simulations

phases/phase0_design/submit_phase0.py --start 1 --end N --dry-run first (preview), then --submit. It (3a) writes per-case scripts via tools/create_case.sh --write-script, (3b) auto-validates via tools/validate_submission_plan.py, (3c.1) builds ONE case fresh (~30 min), (3c.2) runs the rest in --batch-size parallel batches reusing the build's bld/. Non-sequential rounds: --cases-file.

Footgun — build-case reuse. All cases share the build case's bld/. If the build case fails, every dependent case is orphaned — confirm it completed before the batch. Restart it separately, then --skip-build-case --build-case N.

Step 4 — arm monitoring, don't idle-wait

Launch tools/ensemble_auto_monitor.sh (queue polling + auto extraction kick + milestone plots), then hand monitoring setup to arm-hpc-monitoring (CLAUDE.md Rule #6). Point-in-time completion: tools/diagnose_ensemble_status.py --cases 1-N (writes completed/incomplete lists + a validated restart script). Infrastructure failures → restart-failed-jobs; model failures need a fix first.

Step 5 — log it and hand off

Log via calibration-log (phase log → PhaseLogger.log_design): scheme, N cases, param dir, manifest hash, any redesign rationale. Hand off to phase1-exploration once TRANS is >95% complete. (Arrived here from Phase 6 with all candidates pinned at bounds? That's the 6→0 redesign — widen bounds / add parameters and start a new calibration_round.)

Related skills / next phase

Monitoring the in-flight ensemble → arm-hpc-monitoring. Failed jobs → restart-failed-jobs.
Rebuilding a new site/model's knowledge base first → build-rag-from-scratch.
Next: phase1-exploration (extract Y matrix + Morris sensitivity).

Changelog

2026-07-02: Created — offline Phase 0 routine mirroring _run_design(); drives create_parameter_sample → generate_parameter_files → submit_phase0, delegates monitoring/restart, hands off to phase1-exploration.

Mehr aus diesem Repository

gleiches Repository

a2mc-init

jingtao-lbl/A2MC-elm

First-run interactive setup for the offline agent — the official "getting started" flow the FIRST time someone uses A2MC in a repo/site. Interview the user (is a2mc_config.sh set up, where is the E3SM/ELM-FATES checkout, FATES on/off, carbon-only vs nutrient-enabled PARTEH, ECA vs RD, site name/location, PFTs, calibration targets), verify the checkout + RAG milestone, then create and populate the use case (site config + validation targets + parameter list) and hand off to phase0-design. Use when the user says "set up A2MC", "first time using A2MC", "help me get started / onboard me to A2MC", "initialize a new site / use case", "configure A2MC for my site", "I want to calibrate a new site". DISTINCT from onboard-session (which resumes an ALREADY-configured setup).

2026-07-076

add-skill

jingtao-lbl/A2MC-elm

Scaffold + register (or retire + de-register) an A2MC skill so the registry steps are never half-done. Use when the user or agent wants to create a new capability — "add a skill", "create/scaffold a skill", "make this reusable as a skill", "distill X into a skill" — OR to remove one — "remove/retire/delete the X skill". Writes (or deletes) the SKILL.md, and registers (or de-registers) it across all THREE human-facing registries (the README 'Current skills' table, docs/a2mc_reference/skills_catalog.md, AND the AGENTS.md 'At a glance' table), runs the mechanical drift check, and stops for human review before commit. Human-gated.

2026-07-076

arm-hpc-monitoring

jingtao-lbl/A2MC-elm

Set up real-time monitoring of an active A2MC HPC ensemble or experiment on Perlmutter at session start (CLAUDE.md Rule

2026-07-076

calibration-log

jingtao-lbl/A2MC-elm

Log interactive calibration work for a site under use_cases/{site}/memory/logs/. Two log types — a PHASE log (a diagnosis, screening, experiment design/result, refinement — recorded the same way the autonomous agent does, via PhaseLogger, so both modes' logs synthesize together) or a free-form SESSION log (exploratory work that does not map to a specific phase). Use when the user wants to record calibration or exploration work — "log this", "log this phase / diagnosis / experiment", "log this calibration session", "write a session log", "record what I explored".

2026-07-076

compare-calibration-rounds

jingtao-lbl/A2MC-elm

Compare A2MC calibration rounds (R1, R2, …) against each other and the validation targets — top-N best-achievable biomass per round, per-target Morris μ* sensitivity, P-pool / cross-regime overlays. Use when the user asks to "compare rounds", "update the multi-round figure", "refresh the R1-RN comparison", "which round is best", "top sensitive parameters per round", or to regenerate any `multiround_*` / cross-round figure after new cases complete. Codifies the multiround bundle pipeline + the footguns (screening contamination, case-name patterns, μ* ranking, param-set mismatch, partial-ensemble caveats).

2026-07-076

inject-knowledge

jingtao-lbl/A2MC-elm

Inject a human-originated domain fact (a discovery, a parameter insight, a mechanism/relationship) into A2MC's curated knowledge so the calibration agent surfaces it — placing it correctly across the up-to-three channels (site discoveries.json, generic parameters.json, curated_relationships.yaml), validating, and rebuilding the graph. Use when the user says "add this finding/parameter knowledge to A2MC's AI", "inject the X discovery into the KB", "make the agent aware of X", "author a curated relationship for X", or wants a manuscript/literature insight to reach the reasoning pipeline. This is the HUMAN-originated counterpart to curate-knowledge (which promotes run-originated proposals); both are Tier-3 curated writes and you (the human-in-the-loop) own the truth call. Distilled from graphrag_curated_yaml_roadmap.md + dev log 20260519d (the clumping_index injection).

2026-07-076

name	phase0-design
description	Run Phase 0 (DESIGN) of the A2MC calibration workflow as the offline agent — the human-in-the-loop analog of the orchestrator's `_run_design()`. Sample the parameter space (Morris/Sobol/LHS), materialize one FATES parameter file per case, generate + build + submit the ensemble, and arm monitoring. Use when the user says "design a new round", "set up the ensemble", "sample the parameters", "submit the Morris ensemble", "start a calibration round", "expand the parameter space (redesign)".
modes	{"requires_fates":false,"nutrient_pathway":"any","scope":["calibration"],"summary":"Offline analog of online Phase 0 (design/sample/submit the ensemble). Applies in every calibration mode."}

Phase 0: Design & Submit (offline agent)

Floor, not ceiling. This skill is the minimum to stand up a round the way the online agent would. As the offline agent you can do more: reconsider the parameter list itself, add or widen bounds a redesign implies, sanity-check the sampling against prior-round μ* before spending compute, or design a targeted sub-ensemble the fixed loop would never propose. Match the discipline (verify param existence, don't clobber the committed matrix, validate before submit), then exercise judgment about what to sample.

Deliverable: simulation jobs queued (ADSP→RGSP→TRANS dependency chains), submission_manifest.json written, monitoring armed.

Step 1 — sample the parameter space (parameter-file prep — no simulations)

Footgun — matrix overwrite. $A2MC_ENSEMBLE_MATRIX_FILE for Kougarok points at the committed, byte-stable 4890-set matrix the manuscript depends on. To regenerate without clobbering it, write to a scratch path (--output-matrix) or repoint the env var to a fresh round file first.

Step 2 — materialize per-case parameter files (parameter-file prep — no simulations)

Footgun — column↔parameter mapping + file format. The X-matrix column order must match the site parameter list; build_param_lookup() in tools/modify_fates_parameters.py maps each shorthand → FATES name + PFT + organ. Confirm tools/describe_mode.py shows the milestone you expect so the right parameter-file format (JSON vs NetCDF) is written. Never assume a column's meaning.

Step 3 — generate scripts, build, submit the simulations

Footgun — build-case reuse. All cases share the build case's bld/. If the build case fails, every dependent case is orphaned — confirm it completed before the batch. Restart it separately, then --skip-build-case --build-case N.

Step 4 — arm monitoring, don't idle-wait

Step 5 — log it and hand off

Related skills / next phase

Monitoring the in-flight ensemble → arm-hpc-monitoring. Failed jobs → restart-failed-jobs.
Rebuilding a new site/model's knowledge base first → build-rag-from-scratch.
Next: phase1-exploration (extract Y matrix + Morris sensitivity).

Changelog

2026-07-02: Created — offline Phase 0 routine mirroring _run_design(); drives create_parameter_sample → generate_parameter_files → submit_phase0, delegates monitoring/restart, hands off to phase1-exploration.