メニュー
Deep revision history analysis for specific OSM ways — fetches full version history from OSM API v0.6, identifies import bot vs human edits, classifies review status with confidence scores.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Run the full TIGER defect audit pipeline for a MetroNow service zone — fetch from Overpass, classify defects, analyze history, generate reports.
Diff OSM geometry and attributes against TIGER/Line 2024 to identify import drift, new roads, and name-field artifacts in MetroNow zones.
Generate a MapRoulette challenge from scan results for defect classes with high false-positive rates. Creates GeoJSON tasks constrained to MetroNow zone polygons.
Submit corrections to OSM API v0.6 with full community compliance — proper changeset tags, size limits, rate-limit awareness, and dry-run support.
Decompress MetroNow docs so future-you (and fresh AI sessions) can pick the project back up cold without re-deriving everything. Use when the user says docs are confusing, plain, dense, hard to follow, missing context, jargon-heavy, jump straight to conclusions, or are hard to come back to after time away; asks why something is the way it is; says they can't tell what a term means; asks for a walkthrough, explainer, primer, or context refresher; says a section assumes too much; or wants visual aids. Diagnoses missing-middle gaps (undefined jargon, unstated WHY, skipped bridge steps, prose where a picture is needed) and rewrites sections using a fixed template — definition first, then bridge steps, then a load-bearing diagram, then code citations. Outputs to docs/explainers/<topic>.md or in-place edits. No VitePress, no static-site generator — just markdown that GitHub renders.
Unified code review and audit standards for the MetroNow Atlas TIGER Audit Console (https://github.com/AICincy/MetroNow.git). Use this skill when auditing any code in the MetroNow project, doing a PR review, checking code quality, validating accessibility, reviewing Docker config, or when someone asks "review this code," "audit the frontend," "check my PR," "what's wrong with this file," or "does this meet our standards." The codebase is vanilla JavaScript (IIFE pattern, no framework), single-file HTML with inline CSS/JS, IBM Plex typography, Leaflet maps, and an Express.js backend on port 3000 that shells out to the Python `osm` CLI. Covers JS, HTML, CSS, and Dockerfile review with severity-classified findings.
| name | tiger-history-deep |
| description | Deep revision history analysis for specific OSM ways — fetches full version history from OSM API v0.6, identifies import bot vs human edits, classifies review status with confidence scores. |
| when_to_use | User mentions a specific way ID, asks about edit history, checks if a way was reviewed, discusses INCONCLUSIVE status, or wants to investigate a specific street |
| allowed-tools | Read Grep Glob Bash(python *) Bash(curl *) |
| argument-hint | [way-id or street-name] |
| arguments | ["target"] |
Investigate: $target
The core innovation of this project (per Minh Nguyen's feedback) is that tiger:reviewed=no is unreliable. Most mappers don't remove the tag after correcting data. Deep history analysis is the only way to determine if a way has been meaningfully reviewed since the 2007 TIGER import.
out meta data already in scan results)version == 1 and user == "DaveHansenTiger" → definitely UNREVIEWEDGET /api/0.6/way/{id}/history — all versions of the wayGET /api/0.6/node/{id}/history — for nodes along suspicious waystiger:* tags added/removed)?oneway explicitly set by a human or inherited from import?review_status: UNREVIEWED | LIKELY_REVIEWED | INCONCLUSIVEreview_confidence: 0.0 – 1.0These edits do NOT count as human review:
DaveHansenTiger — original 2007-2008 TIGER importsource=tiger_import_dch_v0.6_* tags from nodes~/.config/osm/history_cache/ with 7-day TTL--skip-history flag exists for speed!for d in osm-audit-*/; do f="$d/scan-results.json"; [ -f "$f" ] && echo " ${d%/}: $(python -c "import json; print(json.load(open('$f'))['summary_stats']['total'])") ways"; done 2>/dev/null || echo " (no scans yet)"