Route Researcher — Mountaineering Route Research Skill
Replace 3-5 hours of tab-hopping with one prompt: automatically aggregates 10+ mountaineering sources into a route beta report with weather, avalanche, daylight, and trip reports for North American peaks.
Positives: AskUserQuestion peak confirmation with verifiable links; report template mandates human verification and clearly flags AI-generated content; data gaps explicitly disclosed via a gaps array. Deductions: core dependency is pulled and executed at runtime via uvx from the author's own unverified git repo (dreamiurg/[email protected]), a supply-chain risk; uses Patchright stealth browser to bypass Cloudflare/anti-bot protections, i.e., circumventing third-party access controls; no rollback or isolation guidance.
Positives: layered error handling (partial failure tolerated, gaps array, defined minimum viable report), failures are explained rather than pushed to the user, detailed data contracts and JSON schemas. Deductions: no committed test suite or CI evidence visible in a static read; fetch_conditions.py and cloudscrape.py sources are not in evidence so key paths cannot be confirmed reproducible; cloudscrape.py 'always exits 0' graceful degradation can mask real failures.
Positives: trigger scenarios, example prompts, capability boundaries (North America, dependent on site coverage) and input/output contracts are clearly declared. Deductions: no Chinese-language support evidence; core function depends entirely on overseas services (PeakBagger, WTA, AllTrails, Open-Meteo, NWAC) plus patchright/uvx tooling, so mainland-China reachability is doubtful.
Positives: thorough architecture doc (components, data contracts, design decisions, maintenance instructions), MIT license, version references (v1.10.0, v3.0.0), templates and real example reports, well-layered progressive workflow. Deductions: no CHANGELOG or explicit maintenance ownership/update-path commitment; SKILL.md evidence is truncated mid-Phase-4 so completeness is unconfirmed; Mountaineers MCP is an implicit external dependency whose installation prerequisite is not documented in the evidence.
Positives: two complete example reports (Mount Si, Mount Adams) are highly structured with source links, emergency contacts and permit details; clear marginal value over manual research; templates directly usable. Deductions: examples are author-generated and unverified by third parties; the examples themselves record scraping failures (AllTrails 403), so output completeness varies by target peak.
Positives: example reports cite specific source URLs, dates, authors and word counts; gaps distinguish available from unavailable data; facts and inference are mostly separated. Deductions: static review with no execution evidence and no reproducible tests; all conclusions rest on the files' own claims without multiple independent corroboration.
- The skill pulls and executes code at runtime via uvx from the author's personal unverified git repository (peakbagger-cli), a supply-chain risk; consider hash-pinning or auditing the dependency before use.
- Uses the Patchright stealth browser to bypass Cloudflare and similar anti-bot protections, which may violate target sites' terms of service; assess compliance before enterprise use.
- Core function depends entirely on overseas sites and APIs (PeakBagger, WTA, AllTrails, Open-Meteo, NWAC), likely unreachable or degraded from mainland-China networks; no Chinese-language support.
- Output is AI-generated mountaineering guidance; the template itself requires users to independently verify all critical safety information and is not a substitute for training or judgment.
- Publisher is not verified by FollowSkills; no CHANGELOG or maintenance-ownership commitment, so long-term upkeep is unknown.
What does this skill do, and when should you use it?
Route Researcher is a Claude Code skill that automates mountain route research. Say "Research Mt Baker" and it runs a seven-phase workflow to identify the peak, gather data in parallel, and produce a Markdown route beta report. It aggregates from 10+ sources including PeakBagger, SummitPost, WTA, AllTrails, The Mountaineers, and regional avalanche centers. The hybrid architecture assigns deterministic API data (weather, air quality, daylight, avalanche, hospital and ranger-station lookups) to Python scripts, while three researcher subagents scrape route descriptions and trip reports in parallel. A dedicated reviewer agent validates the final report for factual consistency and safety before delivery.
Searches and confirms the target peak via peakbagger-cli, retrieving coordinates, elevation, and route data; runs fetch_conditions.py to fetch 7-day forecasts, freezing levels, air quality, a full twilight table, avalanche region, ascent statistics, and nearest hospitals/ranger stations/campgrounds; dispatches three parallel researcher agents (PeakBagger+SummitPost, WTA+Mountaineers+regional sites, AllTrails) to pull route descriptions and trip reports, using cloudscrape.py as a fallback ladder for Cloudflare-blocked pages; actively checks road/gate status and permits via state DOT, USFS, NPS, and InciWeb; analyzes route type, crux, and hazards (rockfall, icefall, cornices); then a Report Writer agent generates a templated Markdown report that a Report Reviewer agent validates and fixes.
- A climber planning a Washington volcano who wants weather, avalanche, and road-gate status in one pre-trip report
- A hiker preparing for a classic PNW route like Sahale Peak who needs trip reports and route beta aggregated from multiple sites
- Someone scouting a lightly documented summit to see what route descriptions and recent trip reports exist
- A trip leader who needs a shareable written plan with emergency contacts, ranger-station phones, and a time-table itinerary
- A user wanting a quick natural-language conditions check rather than manual searching, via /mountaineering:conditions in ~30 seconds
What are this skill's strengths and limitations?
- Compresses hours of multi-site manual research into minutes across 10+ specialized sources
- Sensible hybrid architecture: deterministic data via Python scripts saves tokens; judgment tasks go to parallel subagents
- Built-in scraping ladder for Cloudflare protection, and it actively verifies road/gate status instead of telling you to check yourself
- An independent reviewer agent validates dates, math, units, and hazard emphasis; the report carries a prominent AI disclaimer
- Graceful degradation: failed sources are logged in an Information Gaps section rather than silently dropped
- Report quality depends on how well-documented your peak is across the sources; limited for obscure peaks and areas outside North America (the skill itself says it works best for well-documented North American peaks)
- Deeply tied to Claude Code Task subagent orchestration and the Mountaineers MCP; porting elsewhere requires rework
- The Python tools circumvent Cloudflare-style protections to scrape sites, which may violate target-site terms and can break whenever sites change
- Self-showcased example reports carry suspicious future dates (e.g., 2026-01-29), so real-world reliability should be verified by trying it yourself
- Output still requires human verification — the skill itself insists users confirm critical safety info from primary sources
How do you install this skill?
Prerequisites: Claude Code, Node.js (for npx), optionally uv so Python tools install automatically. Recommended: run npx skills add dreamiurg/claude-mountaineering-skills to install the route-researcher skill; flags include -g for global install and --list to preview. If you also want the /mountaineering:* slash commands, install as a plugin instead: in Claude Code run /plugin marketplace add dreamiurg/claude-mountaineering-skills, then /plugin install mountaineering@mountaineering-marketplace, then /reload-plugins.
How do you use this skill?
After install, trigger naturally: "Research Mount Rainier" or "Generate route beta for Forbidden Peak"; or use slash commands: /mountaineering:research <peak> (full report, 3-5 min), /mountaineering:conditions <peak> (weather/avalanche/air quality, ~30 sec), /mountaineering:trip-reports <peak> (recent reports, 1-2 min). Reports save to your current directory as YYYY-MM-DD-peak-name.md. At the tool level, fetch_conditions.py accepts --start-time (adds a trip itinerary) and --waypoint (adds navigation bearings) when you want them.