Skill

wanderlust-goat

Wanderlust GOAT — what a knowledgeable local with great taste would tell you to walk to from here, fused across editorial, local-language, and crowd layers no single tool ranks together.

Verified: 2026-05-13 (printing-press-ingest-2026-05-13+enrich-capability-skill)

When to use wanderlust-goat

Choose if

You're building an agent that recommends what to walk to from a given location, fused across editorial sources (Atlas Obscura, Le Fooding), local-language sources (Tabelog, Naver), and crowd sources (Reddit verbatim quotes). Versus a Google Places search alone, this layers research-plan orchestration, spatial compound queries (food+culture pairs, photo viewpoints with timing), and offline-ready local stores so the agent works even on a spotty connection.

Avoid if

The user's destination is in a region or city where Stage-2 sources lack coverage (the README explicitly flags gaps for lesser-known cities and countries without a Stage-2 region row yet). Also avoid when you need guaranteed-fresh data — Tabelog/Naver/Le Fooding scrapes can be silently intercepted by anti-bot layers, and Reddit quotes are intentionally not summarized. If you have no Google Places key, live discovery won't run.

Risk Flags

  • MEDIUM auth Requires GOOGLE_PLACES_API_KEY for live Stage-1 seeding. ANTHROPIC_API_KEY is optional but materially improves criteria judgment. Without Places key the CLI cannot do live discovery.
  • MEDIUM rate_limit Per-source backoff is implemented but not fully detailed in the README; exit code 7 indicates rate-limited state. Stage-2 sources (Tabelog, Naver, Le Fooding) have anti-bot mitigation that can silently intercept results.
  • MEDIUM scope Stage-2 geographic coverage is uneven — "country X may not have a Stage-2 region row yet". Atlas Obscura slug index doesn't cover every city. Lesser- known cities have known coverage gaps across all regional sources.
  • LOW data_quality Reddit quotes are passed through verbatim (no LLM summarization) explicitly to avoid hallucination. Stage-2 results may be intercepted by anti-bot layers — surface as silent failures unless the agent inspects exit codes.

Cost

Type: Freemium · Free tier: The CLI is free to install. Google Places ships a free $200/month credit per the README (sufficient for moderate use). ANTHROPIC_API_KEY is optional — without it a free heuristic path runs in place of LLM-based criteria judgment. Other regional sources (Atlas Obscura, Tabelog, Naver, Le Fooding) are scraped and cost nothing on the upstream side.

Hidden costs

  • Google Places: $200/month credit covers moderate workloads; heavy use will spill into paid pricing per Google's standard rates.
  • Optional ANTHROPIC_API_KEY usage: token costs per criteria-judgment call when enabled.

Install

Default

npx -y @mvanhorn/printing-press install wanderlust-goat
export GOOGLE_PLACES_API_KEY=<your-key>          # required for Stage-1 live
export ANTHROPIC_API_KEY=<your-key>              # optional: LLM criteria
# Verify external connectivity (Nominatim, OSRM):
<cli> doctor

Setup docs →

Estimated time to first success: ~15 min

Dependencies

Minimum runtime: Node.js (npx), or Go 1.23+, or pre-built binary

Distribution

Repository
https://github.com/mvanhorn/printing-press-library
License
MIT