feat(architecture): restructure page around 5 Alexander principles + framework fixes
Major Changes:
- Restructured architecture.html to lead with Five Architectural Principles
- Added comprehensive 158-line principles section (Not-Separateness, Deep Interlock,
Gradients, Structure-Preserving, Living Process)
- Updated hero section: "Five Architectural Principles for AI Safety"
- Added principle cross-references to Six Services section
- Created i18n documentation for ~50 new translation keys
Framework Improvements (Structure-Preserving):
- Fixed PreToolUse hook to return 'ask' for boundary crossings requiring human judgment
(was incorrectly auto-denying, now properly escalates via Claude Code permission system)
- Fixed CSP validator to only flag NEW violations introduced by edits, not pre-existing ones
(prevents false positives while maintaining security enforcement)
Content Alignment:
- Messaging anchors: "woven into architecture", "critical execution path", "architecturally
impossible to bypass", "evidence-based evolution"
- Cultural DNA compliant (inst_085-089): grounded examples (27027 Incident, inst_064),
evidence-based, anti-consultant, candid about limitations
- Real operational examples: MetacognitiveVerifier selective mode, framework fade detection
Files Changed:
- public/architecture.html: +165 lines (hero + principles section + services intro)
- .claude/hooks/framework-audit-hook.js: Fixed boundary enforcement escalation logic
- scripts/hook-validators/validate-file-edit.js: Improved CSP violation detection
- docs/i18n/ARCHITECTURE_PAGE_I18N_KEYS.md: Translation documentation (new)
Implementation follows SESSION-SUMMARY-2025-10-30-COMPLETE.md action plan (lines 89,
196-237). Structure-preserving transformation: existing content retained, new principles
section adds coherence without breaking existing architecture diagrams or services.
Next Steps: German/French i18n translations needed (~50 keys, 2-3 hours per language)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
8d5edb3126