Delegating a Full Docs Migration to Claude: XOR-Decoding Cloudflare, Finding Hidden Articles, Self-Verifying 22k Words
mhmazur · x · 2026-08-25
An indie developer moved his SaaS support docs from Help Scout into his Rails app, delegating the whole job to Claude. Notable details:
- Hidden docs: after scraping found only 83 articles, Claude probed article IDs 1-170 and recovered 2 lost to pagination plus 2 uncategorized ones — 87 total.
- Cloudflare email obfuscation: it noticed the data-cfemail hex blob's first byte is an XOR key (e.g. 0x0f), decoded help@ by hand, then scripted the rest.
- Resilience: added backoff/retries when throttled; handled malformed HTML (<img>/<p> inside <ol>) and converted everything to Markdown.
- Self-verification: diffed rendered Markdown against live pages — 22,178 words in, 22,178 words out — which is how it caught an empty article; downloaded 190 images behind a CDN.
- Upgrades along the way: restructured URLs with ID-based redirects (which saved 4 stale slug links), recategorized 11 articles, built a search endpoint + UI, and updated all 33 old doc links across 25 files.
A great first-hand case study of an agent handling a real migration end-to-end: task decomposition, anomaly probing, anti-scraping workaround, and a self-check loop.
More from coding & agent
- Agent Tether: Mining Personalized Content with Preference Data — tokenbender · 2026-08-25
- Alibaba treats agent context management as a programming problem — omarsar0 · 2026-08-25
- Skywork 360 pipeline ported to ComfyUI with a custom drone path editor — mickmumpitz · 2026-08-25
- gemini-cli patches SSRF flaw in MCP OAuth metadata discovery flows — josebalius · 2026-08-25
- TrueForge cuts costs via context engineering and on-demand tools — kimmonismus · 2026-08-25
- TrueForge enables local deployment and multi-vendor model support — kimmonismus · 2026-08-25