SkillAgentSearch skills...

arcgis-to-portaljs

Migrate a whole ArcGIS Hub site (opendata.arcgis.com or a Hub Premium custom domain) into a PortalJS Arc portal end-to-end.

Install / Use

npx skills add datopian/portaljs

Installs into whichever agent you are using.

About this skill
⚡

Claude Commands

Claude Code slash commands

Quality Score

85/100

Supported Platforms

Claude Code

Our assessment of arcgis-to-portaljs

arcgis-to-portaljs scores 85/100 on our quality scale, 2612th of 4,700 Development & Engineering skills we index.

Its Claude Commands is 43 KB long, well organised into 30 sections with 10 code examples: long enough that it reads more like full documentation than a focused instruction file, which agents can find harder to follow.

With 2,356 GitHub stars, it is one of the more widely adopted skills in the catalogue.

Substance
21/30
Structure
20/20
Description
15/15
Adoption
14/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated about 2 months ago, so arcgis-to-portaljs is actively maintained.
  • Our last check on 2026-11-04 found the source still online.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-09-30. Automated pattern scan on 2026-09-29. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

arcgis-to-portaljs compared with similar skills

All 4 of these similar skills score higher than arcgis-to-portaljs; compare them before choosing.

SkillScoreStarsUpdatedFormat
arcgis-to-portaljs (this skill)by datopian852.4k2mo agoClaude Commands
claude-memby thedotmack100117.6ktodayCLAUDE.md
taste-skillby Leonxlnx100116.8k8d agoCLAUDE.md
Agent-Reachby Panniantong100112.3k8d agoCLAUDE.md
Understand-Anythingby Egonex-AI10092.8k1d agoCLAUDE.md

Frequently asked questions

How do I install arcgis-to-portaljs?
Run npx skills add datopian/portaljs. The install tabs above show the steps for each supported agent.
Which AI agents does arcgis-to-portaljs work with?
It is written for Claude Code, as a Claude Commands file. Other agents that read the same format can often use it too.
Is arcgis-to-portaljs safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is arcgis-to-portaljs still maintained?
The repository was last updated about 2 months ago, so arcgis-to-portaljs is actively maintained.

description: Migrate a whole ArcGIS Hub site (opendata.arcgis.com or a Hub Premium custom domain) into a PortalJS Arc portal end-to-end. Harvests the Hub /data.json (DCAT-US) inventory, exports every FeatureService layer through the ArcGIS REST query API (resultOffset paging), converts each to the serverless dual tier (PMTiles render + GeoParquet query) with tabular items to Parquet, pushes everything to Cloudflare R2 via Git LFS, appends dual-tier datasets.json entries, and writes a source-vs-derived parity report. The reusable arcgis-to-portaljs migrator. allowed-tools: Read, Write, Edit, Bash(ogr2ogr:), Bash(ogrinfo:), Bash(tippecanoe:), Bash(duckdb:), Bash(git:), Bash(curl:), Bash(jq:), Bash(npx:), Bash(node:), Bash(mkdir:), Bash(cp:), Bash(wc:), Bash(zip:), Bash(command:), WebFetch

/arcgis-to-portaljs

Migrate an entire ArcGIS Hub open-data site into a PortalJS Arc portal in one pass. Every Hub site is machine-readable — a DCAT-US catalog at /data.json, and every dataset behind it is an ArcGIS REST FeatureService with a documented query API — so migration is a harvest → export → convert → publish → verify pipeline that is almost fully automated. The tooling is the reusable arcgis-to-portaljs migrator: input = a Hub URL, output = a ready-to-deploy PortalJS catalog + a parity report. This is the repeatable asset, not a one-off.

Input is one Hub URL (e.g. https://hub-lewisville.opendata.arcgis.com or https://streamwaterdata.co.uk). Output: datasets.json entries, data on R2, and arcgis-parity-report.md.

What this composes — do NOT reinvent

This skill is the ArcGIS-Hub-specialized orchestrator. It reuses three existing pieces and adds two ArcGIS-specific steps (the REST export loop and the parity report):

| Step | Owned by | This skill's role | |------|----------|-------------------| | Harvest /data.json (DCAT-US → canonical) | /portaljs-migrate (dcat source) | Reuse its DCAT-US field map; add Hub item classification (FeatureService vs table vs non-data). | | Export FeatureService → GeoJSON (REST paging) | new here | /portaljs-migrate's arcgis mode only links the query URL; this downloads the features so they can be converted and served serverless. | | Convert → PMTiles + GeoParquet (dual tier) | /portaljs-add-geo | Reuse its exact ogr2ogr/tippecanoe/duckdb recipes per layer. | | Publish: bulk Git-LFS → R2 + datasets.json | /portaljs-migrate (download mode) + /portaljs-add-geo (§10 dual-tier entry) | Reuse the bulk Giftless push; emit dual-tier entries. | | Parity report (source vs derived) | new here | Record counts, bbox, attribute schema, geometry validity per layer. |

Naming note (Phase 0, for review): the bead names the deliverable arcgis-to-portaljs (a directional, pitch-friendly asset name). The rest of the suite is prefixed portaljs-. Renaming to portaljs-migrate-arcgis for suite consistency is a trivial follow-up if the reviewer prefers it — the skeleton is name-agnostic.

Scope — Phase 0 skeleton

Per the migration epic (po-0qe §6), Phase 0 delivers the automatable core — the pipeline above, proven runnable. Explicitly out of scope here (tracked as child beads under po-0qe, do not build in this pass):

  • Content pages / branding (Hub layout pages → MDX) — Phase 2 (Stream).
  • Multi-publisher org structure (one namespace/repo per publisher) — Phase 2 (Stream).
  • Scheduled sync + cutover playbook (re-harvest on cadence, redirect map, DNS flip) — Phase 3.
  • Koop FeatureServer-compat adapter (keep old REST consumers working) — Phase 4 (optional).
  • 3D scenes / StoryMaps / drone imagery — no serverless open equivalent; link-out only.

Required input — ask, don't error

  • Hub URL (required) — the site root, e.g. https://hub-lewisville.opendata.arcgis.com or a Hub Premium custom domain (https://streamwaterdata.co.uk). The skill appends /data.json itself; accept either the root or a full /data.json URL.
  • Portal directory — path to the target PortalJS portal (default: current directory). If none exists, tell the user to run /portaljs-new-portal first.
  • Project slug — the Arc deploy slug for the R2 push (same slug /portaljs-deploy uses).

Optional flags: --limit N (cap datasets, for a trial run), --only <slug,slug> (migrate a subset by Hub slug), --dry-run (harvest + plan + print, write nothing), --namespace-mode theme|owner (default owner — one namespace per Hub publisher; see /portaljs-migrate).

If the Hub URL is missing, ask for it — never dead-end. Example:

I need an ArcGIS Hub site URL to migrate. Provide the site root, e.g.
  https://hub-lewisville.opendata.arcgis.com
(a Hub Premium custom domain like https://streamwaterdata.co.uk also works.)

Runtime — native shell-out

Same native stack as /portaljs-add-geo (GDAL, tippecanoe, duckdb+spatial) plus jq for JSON slicing and node/curl for the REST paging. No server, no container — the whole migration runs on the operator's machine, keeping PortalJS's zero-backend bet.

Steps

1. Gather input (interview if thin)

From $ARGUMENTS extract HUB_URL, PORTAL_DIR (default .), PROJECT_SLUG, and the optional flags. Normalize HUB_URL: strip a trailing /data.json and any trailing slash to get SITE_ROOT; DATA_JSON="$SITE_ROOT/data.json". If HUB_URL is absent, emit the prompt above and stop.

2. Check native tools (hard gate)

Delegate to /portaljs-add-geo §2 — detect ogr2ogr, tippecanoe, duckdb (+ spatial), and additionally jq. Any missing → print the per-OS install and stop with ERROR: [arcgis-to-portaljs] MISSING_TOOLS <tools>:

for t in ogr2ogr tippecanoe duckdb jq node; do
  command -v "$t" >/dev/null 2>&1 || MISSING="$MISSING $t"
done
# macOS:  brew install gdal tippecanoe duckdb jq node
# Debian: sudo apt-get install -y gdal-bin duckdb jq nodejs   # tippecanoe: build from source
duckdb -c "INSTALL spatial; LOAD spatial; SELECT 1;" >/dev/null 2>&1 || MISSING="$MISSING duckdb-spatial"

3. Validate the portal directory

Confirm PORTAL_DIR is a PortalJS portal: datasets.json (a JSON array) and the geo showcase components (components/MapPreview.tsx, components/GeoQuery.tsx, PR #1647+) exist. Missing → ERROR: [arcgis-to-portaljs] NOT_A_PORTAL <dir> — run /portaljs-new-portal first, then re-run. Missing geo components → tell the user to update the template (the dual tier needs them).

4. Harvest the Hub inventory (/data.json → canonical + classification)

One GET of the DCAT-US feed; datasets live under dataset[]. Reuse /portaljs-migrate's DCAT-US mapping verbatim (title/description/keywords/publisher/temporal/spatial map ~1:1) to build the canonical list. Then classify each item by its distributions — this is the Hub-specific step DCAT alone doesn't give you:

curl -fsSL "$DATA_JSON" -o /tmp/a2p-datajson.json
jq '.dataset | length' /tmp/a2p-datajson.json   # inventory size

Classify each dataset[] item:

| Class | Detect on | Migration path | |-------|-----------|----------------| | vector | a distribution accessURL matching /FeatureServer/\d+$ or /MapServer/\d+$ (note the trailing layer index — service-root URLs with no /N are web maps, not data) and the title is not a 3D/mesh/tile give-away | REST export (§5) → dual tier (§6) | | table | only tabular distributions (CSV/Excel), no geometry | download → Parquet (§6b) | | non-data | everything else: web maps (a REST distribution at the service root, no /N), 3D scenes/meshes/multipatch, tiles.arcgis.com/.../MapServer tile services, dashboards, StoryMaps, web apps, imagery, Esri basemaps | skip — log to the report as "link-out / out of scope", do not migrate |

/data.json does not carry an Esri item type. DCAT-US gives you title, description, and distributions — not the AGOL item type (Web Map / Web Scene / Dashboard …). So classify on distribution shape, not a type field: the presence of a real …/FeatureServer/<id> (with the numeric layer index) is the signal that data can be exported. A distribution whose accessURL stops at the service root, or points at tiles.arcgis.com, or whose title screams 3D/mesh/LiDAR-colored/reprojected, is a viewer or tile cache — skip it. Then probe each vector candidate (?f=json → geometryType, /query?returnCountOnly=true → count): a real feature layer returns a point/line/polygon geometry type and a count; a 3D/multipatch or broken service does not. This probe is what separates a genuine vector layer from a REST-shaped non-data item. (Lewisville: 44 items → 7 real vector layers, 37 skipped.)

Record per item: the FeatureServer layer URL (strip to …/FeatureServer/<id>), geometry type, format list offered, issued/modified (source dates — see §4b; modified also drives future sync, Phase 3), and publisher (the namespace under --namespace-mode owner).

Multi-publisher namespace resolution (--namespace-mode owner, Phase 2/Stream). Raw publisher.name strings are NOT namespace-safe on a multi-publisher Hub — Stream's feed carried 28 labels for ~22 real publishers (Anglian Water + Anglian Water Services, Yorkshire Water + Yorkshire Water Services, Northumbrian Water + … Ltd) plus 53 items with the literal broken template {{source}}. Resolve in three steps: (1) a normalization table raw label → canonical namespace slug (build it from the feed's publisher histogram; merge obvious legal-suffix variants); (2) for {{source}}/empty publishers, longest title-prefix match against the canonical brand list — include sub-brands mapped to their parent (Essex And Suffolk Water → northumbrian-water, Hartlepool Water → anglian-water); (3) fallback to the platform's own namespace (e.g. stream). Keep a namespace → display-title map for the catalog. Don't derive namespaces from Hub <org>::<slug> landing-page keys — on Stream only 29/378 items carried one. Dedup near-duplicate view layers: a Hub often exposes the same underlying data twice — a hosted view + its source layer (Lewisville: "Monument Survey Network Benchmarks" AND "… Benchmarks view", two identical 39-point items). Both as separate catalog entries is source junk, not fidelity. Detect: normalize each title (lowercase, collapse non-alphanumerics to single spaces, strip a trailing view token) — two vector items that collide on the normalized title are a dedup candidate; confirm with the §5 record counts (equal ⇒ same data). Migrate only the source layer (the one whose raw title has no view marker; if both or neither carry it, keep the later modified) and skip the other, logging the pair to the report as deduped: <kept> ⇠ <skipped>. Merge metadata across the pair: the twins often carry different descriptions/snippets (Lewisville: the view had the clean one-line summary, the source layer a vendor contact block) — run both through the §4b description chain and keep the better survivor (a short human summary beats an address/contact blob) on the kept entry. If counts differ, the "view" is a real filtered subset — keep both and only flag. On re-runs the upsert must also remove a previously migrated entry that this rule now skips (§7). The count check is MANDATORY before any drop, not a confirmation nicety (Phase 2 lesson): on a multi-publisher Hub, different datasets legitimately share a normalized title within one namespace (Stream/Wessex exposed five Asset Nodes / Model Links / Flow Pressure 15min sets — different scenario runs, different counts). Probe returnCountOnly for BOTH twins; equal ⇒ dedup, different ⇒ keep both (suffix the slug -2, -3, …) and log them as restored-distinct, never deduped. A title-only dedup on Stream would have silently dropped 24 real datasets. Cross-check completeness:

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars2.4k
CategoryDevelopment
Updated2mo ago
Forks335

Languages

TypeScript

Trust signals

100/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

No cautions