docs: bring wiki current with v0.4.0
Updates every page for the security layer, the five indexability-contradiction
gates, Google rich-result validation, on-page accessibility gates, and the
per-layer scoring model that replaced the flat 100-point budget. Regenerates
Finding-Reference from the registry (48 -> 67 ids), recounts the failOn enum
(28 -> 42 gate names), and adds Security-Layer and Contradictions with a real,
verified true positive against danluu.com. Removes every version-numbered
--write promise in favour of naming the condition it ships under.
docs: bring wiki current with v0.3.0
Every page predated v0.2.0 and none mentioned omnirank fix. Update all 13
existing content pages plus the sidebar and footer for the 0.2.0 site-level
gates/perf layer/script-aware AnswerBlock bands, the 0.2.1 false-positive
fixes/GATE_CAP/notEvaluated/geo.license default change, and the 0.3.0
locator/fixTier/applicability/fix preview. Add four new pages: Fix-Preview,
Fix-Tiers-and-Applicability, The-Locator, and a Finding-Reference table
generated directly from registry.py so it cannot drift from the code.
docs: publish the full OmniRank wiki
Fifteen pages covering quick start, both shipped skills, configuration and
report schemas, CI recipes, Claude Code setup, troubleshooting, a 32-term
glossary, a 20-question FAQ, the research this project builds on, and the
roadmap.
Wiki pages are search-indexable, so they follow the practices the tool itself
enforces: one H1 per page, a 40-60 word answer-first opening, question-shaped
headings, and tables for reference material.
Research-and-Evidence states plainly that the cited correlations are
observational, drawn largely from English-language B2B datasets, and are
directional rather than guaranteed.