heavymap

Cursor cloud environment (E3 atlas)

Status: starter, 2026-10-16. The repo routines are still being decided, so this setup is deliberately thin.

Piece	Value
Repos	niagaraassembly/heavymap only (industrial-atlas later, if it becomes a repo)
Build	bash scripts/cloud-install.sh: Python check, byte-compile, unit tests, planning-index path check. No installs (stdlib only), no data pulls
Start	bash scripts/cloud-start.sh: nothing (no server; nothing published before G6)
Network	set in .cursor/environment.json: egressMode: network_settings_only + egressAllowlist (GitHub hosts, PyPI); add open-data hosts there once named
Secrets	none

Open questions tied to routines still being decided

1. Data pulls from a cloud agent. DATASET-ROUTINE.md keeps pulls in local local-data/. A cloud VM is disposable, so profiling there loses the raw pull. Allow cloud profiling (record facts only, then discard), or keep pulls to deb/local agents?
2. Open-data allowlist. Which publisher hosts (ArcGIS servers, portals) go on the environment allowlist? Comes from Spreadsheets/16-service-profile.csv once services are profiled.
3. Draft skill. heavymap-planning/drafts/dataset-routine-SKILL.md is unregistered; .agents/skills/heavymap-dataset-profiling only points at it. Promote it once the routine is approved (and settle defer vs needs_more, OPEN-DECISIONS 1).
4. MCS authority (OPEN-DECISIONS 5): may cloud agents edit Spreadsheets/*.csv at all, or only propose?
5. hm/fresh-foundation (12 commits, CLI scripts/hm, registry, more tests). If it lands, cloud-install.sh should run its tests and any requirements it adds.
6. CI. There is no GitHub workflow; should the same checks run on PRs?
7. Lane. Cloud agents take agent:cloud issues only (AGENTS.md); confirm routing rule Industrial Atlas → E3 should require that label.