Role-indexed translation layers for Claude's capabilities — written in your professional vocabulary.
Live site: archiecur.github.io/doc_translation_layer
The Claude documentation is written for developers. This site is written for everyone else.
Each page in this collection is a translation layer — a role-indexed guide that maps Claude's capabilities to the problems, fears, and professional standards of a specific role. Not a feature list. Not a tutorial. A door into Claude's capabilities, labeled in your vocabulary and described in terms of what it does for your work.
Current roles:
- Auditor — Control frameworks, evidence standards, audit trails
- Marketer — Brand integrity, compliance exposure, asset protection
Each capability entry has two views: one for the professional, one for the developer implementing it. Both are served by the same entry. Every claim links directly to Anthropic's documentation.
Three people are often in the same room when an organization decides whether to deploy an AI system: the professional who owns the risk (an Auditor, a Legal officer, a compliance lead), the developer who builds the controls, and the business owner who needs it to work.
Right now, they are having three different conversations, in three different vocabularies, with no common ground to stand on. The developer cannot explain the controls in language the risk owner recognizes. The business owner cannot demonstrate to Legal that the system is compliant.
This site gives all three of them the same verified reference. The claims are grounded in Anthropic's documentation — followable by anyone in the room. The Auditor can cite the entry. The developer can point to the implementation reference. The CISO can follow the link and verify the control herself.
The Compare view makes the underlying structure visible: the same Claude capability — hooks, structured outputs, tool allowlists — appears across multiple roles because it solves a fundamentally different problem for each of them. The Auditor needs structured outputs because she cannot verify what she cannot reproduce. The Marketer needs structured outputs because brand drift happens at volume. Same capability, different reason for walking through the door.
This is a shared reference architecture for organizational AI adoption conversations — not just helpful to individual roles, but to the governance layer that has to sign off on deployment.
Each role lives in a single markdown file in _roles/. Adding Role 3 means creating one file. No layouts, templates, or JavaScript are touched.
See CONTRIBUTING.md for the full content operation, front matter schema, and capability ID convention.
Requires Ruby and Bundler.
bundle install
bundle exec jekyll serve
Site runs at http://127.0.0.1:4000/doc_translation_layer/
MIT License — see LICENSE
ArchieCur · Sonnet 4.6 (Anthropic) · Claude Code · 2026