laurigates/claude-plugins
Overview
This skill derives Architecture Decision Records (ADRs) for an existing project by analyzing its code structure, dependencies, and documentation. It automates discovery, conflict analysis, and MADR-style ADR generation so implicit architecture choices become recorded and actionable. The output includes ADR files, an ADR README, and manifest updates for Blueprint Development.
How this skill works
The skill inspects project files (directory layout, package manifests, config, imports, test folders) and runs an Explore agent to surface frameworks, data layers, API styles, and other patterns. It checks existing ADRs for conflicts, calculates conflict scores, prompts for relationships, and generates MADR-formatted ADRs plus an ADR README and manifest updates. Superseded ADRs are updated bidirectionally and a summary report is produced.
When to use it
- Onboarding an existing codebase into the Blueprint Development system
- Documenting implicit or legacy architecture decisions for traceability
- Preparing architecture artifacts before major refactors or handoffs
- When you want a programmatic ADR index and consistent manifest entries
- After running /blueprint:derive-prd (recommended) or when PRD is missing
Best practices
- Run the discovery phase with the Explore agent to capture evidence from code and configs
- Run /blueprint:derive-prd first if a PRD exists to link requirements to ADRs
- Resolve high conflict-score ADRs interactively: supersede, extend, or relate
- Document only decisions with real alternatives; mark unclear rationales as 'needs rationale'
- Keep ADRs concise and focused on 'why' rather than implementation details
Example use cases
- Generating ADRs for language, framework, and testing choices in a migrated repo
- Detecting and resolving conflicting database or deployment decisions across ADRs
- Producing an ADR README that auto-indexes records for team reference
- Marking inherited decisions as legacy and flagging items needing team rationale
- Creating a set of standard ADRs (0001–0008) for a monorepo during onboarding
FAQ
The project should have docs/blueprint/ present. A PRD is recommended but not required; if missing the skill can prompt to run /blueprint:derive-prd first.
How does the skill handle existing ADR conflicts?
It scores potential conflicts, surfaces those with high scores, and prompts you to supersede, extend, relate, or ignore each conflicting ADR. Superseded ADRs are updated for bidirectional consistency.
Which ADRs does the skill create by default?
If applicable, it creates standard ADRs such as project-language, framework-choice, testing-strategy, styling-approach, state-management, database-choice, api-design, and deployment-strategy.
8 skills
This skill derives Architecture Decision Records from an existing project by analyzing code structure, dependencies, and documentation to onboard Blueprint.
This skill helps you process YAML efficiently with yq, including querying, filtering, and transforming manifests and configurations.
This skill helps you write and run Python tests with pytest, coverage, fixtures, and mocks to improve reliability.
This skill helps Rust projects run tests faster and more reliably by isolating tests in separate processes with flexible filtering.
This skill generates architecture diagrams and flowcharts from declarative D2 text, enabling rich styling, layouts, and themes with automatic rendering.
This skill helps you query, filter, and transform JSON using jq to extract insights and reshape data for workflows.
This skill helps you enforce Python code quality by configuring and applying ruff and mypy across projects for linting, formatting, and type checking.
This skill helps you master modern Python development with type hints, async patterns, and Pythonic idioms for reliable, readable code.
