Skip to content

What is pentaglyph?

pentaglyph is a documentation scaffold — a kit of templates, directory layouts, and a single canonical workflow — that binds five industry-standard practices into one project structure your humans and your AI agents can both navigate.

The name comes from Greek penta (“five”) + glyph (“engraved sign”). Five peer standards, engraved into one opinionated layout, plus a sixth slot (the Project Engagement Layer) that composes eight well-known client-communication primitives for consulting work.

#StandardWhat it doesAuthoritative source
1arc42Architecture documentation — 12-section templatearc42.org
2C4 modelVisual notation for software architecturec4model.com
3MADR v3.0Markdown Architecture Decision Recordsadr.github.io/madr
4DiátaxisFour-quadrant technical writing modeldiataxis.fr
5TiSDDService design methods (personas, journeys, blueprints)thisisservicedesigndoing.com/methods
6PEL (binder)8 client-engagement primitives composed under one homeper-primitive URLs in template/docs/STRATEGY.md §2.6

External standards stay authoritative. pentaglyph does not invent a sixth peer standard — slot 6 is a binder that composes existing primitives (Inception Deck, GitLab Handbook weekly update, Atlassian weekly status, Basecamp Heartbeat, Amazon 6-pager, Now-Next-Later, DACI, RAID, PR-FAQ).

Standards alone are not a system. Most teams adopt arc42 or Diátaxis or MADR in isolation and end up with three half-followed conventions, no agreed lifecycle, and AI agents that have no idea which directory to write to. pentaglyph adds:

  1. A concrete file layout that maps each standard to a directory under docs/ — predictable, hard to misplace.
  2. A single canonical workflow (docs/WORKFLOW.md) that tells humans and AI agents when to write what, where to put it, and what lifecycle state it goes through (Draft → Active → Superseded → Archived).
  3. Per-directory README.md files with explicit AI instructions so an LLM can place new content correctly with zero project context.
  4. A Bun-based CLI (cli/) that scaffolds a new project’s docs/ from this template with profile, language, and AI-target options.
Terminal window
bunx --bun @uyuutosa/pentaglyph init ./my-project --profile=standard --ai=claude

That command creates:

  • ./my-project/docs/ — populated with the full kit (all five standards’ templates).
  • ./my-project/.claude/rules/documentation.md — the auto-load rule for Claude Code.

Then open ./my-project/docs/AI_INSTRUCTIONS.md and ./my-project/docs/WORKFLOW.md. Those two files contain everything you need.

  • Engineering teams that want to standardise architecture, decision, and user-facing documentation without writing their own scaffold.
  • AI-augmented teams (Claude Code, Cursor, Copilot Workspace) that need documentation an agent can navigate and update without human babysitting.
  • Consulting / advisory practices that need client-engagement artifacts (status updates, decision logs, 6-pagers) co-located with technical docs.
  • Teams that already have a mature internal doc system they are happy with.
  • Single-file projects where README.md is sufficient.
  • Teams that want a single proprietary standard instead of composing established open ones — pentaglyph deliberately defers to upstream sources.
  • 30-minute tutorial — PRD → ADR → Module DD → code with paired doc. The fastest way to feel how the kit clicks together.
  • Why pentaglyph exists — the design rationale and what gap in the documentation-standards ecosystem it fills.
  • Use with Claude Code — the AI-agent workflow, including the auto-load rule and prompt patterns.