spedas-workflow

SPEDAS Agent Kit workflow

The plugin exposes one MCP server named spedas. Start with spedas_overview() when uncertain.

Prefer the public SPEDAS mental model:

1. science workflow layer;
2. unified data layer;
3. data source categories: cdaweb, pds, spice;
4. internal backend packages only when maintaining/debugging the MCP.

Preferred tools

• search_spedas_data_sources
• plan_spedas_observation
• compare_cdaweb_pds_spice
• create_spedas_analysis_bundle
• browse_data_sources(source_type="all"|"cdaweb"|"pds"|"spice")
• load_data_source(source_type, source_id)
• browse_data_parameters(source_type, dataset_id, ...)
• fetch_data_product(source_type, ...)
• manage_data_cache(source_type, ...)

Compatibility low-level tools remain available for maintenance/debugging, but new agent workflows should start with the unified data-layer tools.

MMS reconnection events (Batch 006 guardrail)

For MMS reconnection/EDR papers, keep the first pass narrow and explicit: plan with spedas_overview / plan_spedas_observation, fetch burst FGM/FPI/EDP artifacts, then chain into existing analysis helpers before asking for new code. Use analyze_minvar_coordinates plus transform_timeseries_coordinates for LMN or field-aligned panels, and use *-DIST artifacts with compute_particle_spectra(..., spectrum_types=["energy", "pitch_angle"]) for PAD/energy claims. A single-spacecraft e*n_e*(V_i-V_e) current or J·E' is a transparent proxy, not a curlometer or paper-quality heating result; mark it proxy unless the interval, LMN/FAC basis, calibrated E-field, and MMS1-4 curlometer diagnostics are all verified. If the paper/supplement interval cannot be verified, record candidate_interval or availability_failure instead of widening the fetch or claiming reproduction.

Heliospheric ICME/SEP multi-spacecraft events (Batch 007 guardrail)

For Wind/ACE/STEREO/OMNI ICME, magnetic-cloud, CME-CME, or SEP papers, use paper-reproduction as the outer artifact contract and treat docs/examples/stereo_icme_multispacecraft.md as the reduced first-pass recipe. Start with STEREO MAG 1min + PLASTIC proton moments for multi-day events, add Wind/ACE/OMNI only with explicit source/propagation labels, and keep SEP products as reduced_sep_proxy until telescope/species/energy-channel metadata are verified. Batch 007 confirmed that STEREO/PLASTIC/SEPT routing already exists; the repeated gap is discoverability, event seeds, variable-alias provenance, and overclaim prevention. Do not promote shock/sheath/cloud boundaries, SEP onset/fluence, or SECCHI/HI J-map context to paper-quality outputs unless those products and methods are explicitly loaded and documented.