Doc Architect

Platform Notes

• Optional helper plugins may help in some environments, but they must not be treated as required for this skill.

Doc Architect

Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.

Use When

• Generate Triple-Layer AGENTS.md documentation by scanning a project for its tech stack, data directory, and planning directory. Use when the user asks to standardize project documentation, generate agent files, or create AGENTS.md guides.
• The task needs reusable judgment, domain constraints, or a proven workflow rather than ad hoc advice.

Do Not Use When

• The task is unrelated to doc-architect or would be better handled by a more specific companion skill.
• The request only needs a trivial answer and none of this skill's constraints or references materially help.

Documentation Intake Guidance

• Gather relevant project context, constraints, and the concrete problem to solve; load references, templates, protocols only as needed.
• Confirm the desired deliverable: design, code, review, migration plan, audit, or documentation.

Documentation Method Detail

• Read this SKILL.md first, then load only the referenced deep-dive files that are necessary for the task.
• Apply the ordered guidance, checklists, and decision rules in this skill instead of cherry-picking isolated snippets.
• Produce the deliverable with assumptions, risks, and follow-up work made explicit when they matter.

Quality Standards

• Keep outputs execution-oriented, concise, and aligned with the repository's baseline engineering standards.
• Preserve compatibility with existing project conventions unless the skill explicitly requires a stronger standard.
• Prefer deterministic, reviewable steps over vague advice or tool-specific magic.

Legacy Documentation Pitfalls

• Treating examples as copy-paste truth without checking fit, constraints, or failure modes.
• Loading every reference file by default instead of using progressive disclosure.

Documentation Deliverable Detail

• A concrete result that fits the task: implementation guidance, review findings, architecture decisions, templates, or generated artifacts.
• Clear assumptions, tradeoffs, or unresolved gaps when the task cannot be completed from available context alone.
• References used, companion skills, or follow-up actions when they materially improve execution.

Evidence Produced

References

• Use the references/ directory for deep detail after reading the core workflow below.
• Use the templates/ directory when the task needs a structured deliverable.
• Use the protocols/ directory for formal execution order or handoff rules.

Design and generate a portable Triple-Layer AGENTS.md documentation set that reflects the project’s real structure and constraints.

When generated documentation must describe system boundaries, workflows, requirements, data architecture, interfaces, states, business rules, or design-system conventions, load systems-process-requirements and use its artifact router before writing the docs.

Modularize Instructions (Token Economy): Avoid consolidating all AI/dev guidance into a single CLAUDE.md. Prefer smaller, focused docs (e.g., docs/setup.md, docs/api.md, docs/workflows.md) and reference them only when needed.

Documentation Standards (MANDATORY): ALL generated markdown files must follow strict formatting rules:

• 500-line hard limit - no exceptions for any .md file
• Two-tier structure: High-level TOC (Tier 1) + Deep dive docs (Tier 2)
• Smart subdirectory grouping for related documentation
• See doc-standards.md for complete requirements

Core Outcome

Inputs

Capability Contract

Inspection is read-only by default. Creating directories or changing instruction files requires explicit authority; deletion and unrelated documentation changes are out of scope.

Documentation Fallback Notes

Without full workspace access, produce a partial documented map with unknown areas named; do not claim the three layers are complete.

Decision Rules

Documentation Correction Examples

• Inventing a technology stack; inspect manifests.
• Copying generic rules; bind them to evidence.
• Placing files arbitrarily; use discovered boundaries.
• Editing without authority; return a plan.
• Hiding conflicting instructions; escalate them.

Worked Example

If no schema directory exists, the output proposes a data-layer location and records the gap instead of creating one silently.

Core Outcome

Produce three aligned AGENTS.md files:

• Root AGENTS.md: Project identity, tech stack, global standards
• Data AGENTS.md: Data integrity rules and schema governance
• Planning AGENTS.md: Spec-driven development workflow

Trigger Phrases

The skill should activate when the user asks to:

• Standardize project documentation
• Generate agent files

Standard Operating Procedure (SOP)

1. Scan the workspace

   • Inspect the root for identifiers (README, PROJECT_BRIEF, TECH_STACK, ARCHITECTURE, CLAUDE, package.json, composer.json, *.sln, pyproject.toml).
   • Locate likely data directories (database/, schema/, migrations/, sql/, db/).
   • Locate planning/documentation directories (docs/, docs/plans/, planning/, specs/).
   • Identify module/area entry points (menus, docs, feature folders) to group specs.
   • Note template conventions (public/ as web root, per-panel includes, API outside public).

2. Identify the environment

   • Determine primary language (PHP/C#/Python or other).
   • Determine DB type (MySQL/PostgreSQL/SQLite/SQL Server/other).
   • Determine deployment environment (Docker/Kubernetes/shared hosting/cloud).

3. Set up plan grouping (first-time)

   • Create docs/plans/<module>/ subdirectories for each discovered module/area.
   • Update docs/plans/AGENTS.md with the current module list.
   • Create or update docs/plans/INDEX.md as the master plan status index.
   • Ensure the index includes status, urgency, last implementation date, and last modification date.
   • Keep docs/plans/AGENTS.md updated whenever plans are added or their status changes.
   • Maintain a folder map at the top of docs/plans/AGENTS.md and update it when requested.
   • Note that developers can add new folders and update the list manually.

4. Generate Triple-Layer docs

   • Use the templates in templates/root-agents.md.template, templates/data-agents.md.template, and templates/plan-agents.md.template.
   • Populate with real findings and pull constraints from references/logic-library.md as needed.
   • Create files at:
     • Root: AGENTS.md at project root
     • Data: database/schema/AGENTS.md (or best-fit schema directory)
     • Planning: docs/plans/AGENTS.md (or best-fit planning directory)

Bundled Resources

• protocols/workflow.md: 3-step workflow used during generation
• templates/root-agents.md.template: Root AGENTS.md template
• templates/data-agents.md.template: Data AGENTS.md template
• templates/plan-agents.md.template: Planning AGENTS.md template
• references/logic-library.md: Domain constraint library for reuse

Common Pitfalls

• Do not invent tech stacks. Only infer from files found in the workspace.
• Do not place AGENTS.md in arbitrary locations; follow the best-fit paths above.
• Do not include contradictory rules across the three layers.

Quick Example

If a project uses Laravel + MySQL with docs/plans and database/schema:

• Root: AGENTS.md → PHP/Laravel, MySQL, deployment standards
• Data: database/schema/AGENTS.md → referential integrity, no-delete rules
• Plans: docs/plans/AGENTS.md → spec.md format and workflow steps

Cross-References to SDLC Skills

When generating AGENTS.md files, be aware of the complete SDLC documentation ecosystem:

SDLC Documentation Skills

Related Documentation Skills

SDLC Output Directory Structure

When scanning for documentation, expect this structure in projects using SDLC skills:

docs/
├── planning/        # sdlc-planning output (7 docs)
├── design/          # sdlc-design output (6 docs)
├── testing/         # sdlc-testing output (5 docs)
├── user-deploy/     # sdlc-user-deploy output (6 docs)
├── plans/           # feature-planning output
│   ├── AGENTS.md    # Plans directory index (doc-architect manages this)
│   ├── INDEX.md     # Plan status tracker
│   └── specs/       # Feature specifications
└── project-requirements/  # project-requirements output

Integration Rule: When generating the Planning AGENTS.md (docs/plans/AGENTS.md), include references to any SDLC documentation directories that exist alongside the plans directory.

Workflow

1. Inspect manifests, instructions, data paths, planning paths, and boundaries read-only.
2. Map evidence to root, data, and planning instruction layers.
3. Stop when authority is absent, instructions conflict, or a location lacks evidence.
4. Recover with a qualified plan, mark unassessed areas, and request resolution before mutation.
5. Generate authorised files and verify consistency, links, placement, and line limits.

Outputs

Degraded Mode

Without full workspace access, return a qualified partial map, preserve unassessed directories, and do not claim the three-layer set is complete.

Anti-Patterns

• Inventing a technology stack. Fix: inspect manifests and source files.
• Copying generic rules. Fix: bind each rule to repository evidence.
• Placing files arbitrarily. Fix: use responsibility boundaries.
• Editing without authority. Fix: return a proposed file plan.
• Hiding conflicting instructions. Fix: stop and request resolution.