Summary
Add a skill for proactive architectural friction discovery and refactoring planning.
Problem / Motivation
Forge has no skill for identifying architectural problems before they become bugs. Current workflow is reactive: issues are created when problems are noticed. A proactive skill would explore the codebase organically, surface friction, and propose targeted refactors.
Inspired by mattpocock/skills improve-codebase-architecture, which combines organic exploration with structured dependency analysis.
Proposed Solution
Create forge-improve-architecture with these core concepts:
Organic Codebase Exploration
- Explore like a new team member — note where understanding requires bouncing between many files
- Surface friction: shallow modules, tightly-coupled seams, untested code, functions extracted just for testability
Dependency Categories Framework
Classify each friction point by dependency type, which determines the testing strategy:
- In-process — pure computation, no I/O → merge and test directly
- Local-substitutable — has test stand-ins (e.g., SQLite for Postgres) → deepen with local sub
- Remote but owned (Ports & Adapters) — own services across network → define port, inject adapters
- True external (Mock) — third-party services → mock at boundary
Parallel Sub-Agent Design Exploration
- Spawn 3+ sub-agents with radically different design constraints
- Each proposes a different interface for the identified friction point
- Compare on: simplicity, general-purpose vs specialized, depth, ease of correct use
"Replace, Don't Layer" Testing
- Delete old tests on shallow modules once boundary tests exist
- Write new tests at the deepened module's interface boundary
- Assert on observable outcomes through public interface
Pipeline Fit
Standalone skill invoked ad-hoc — not part of the linear workflow. Can feed into forge-create-issue when a refactor is identified.
Acceptance Criteria
Summary
Add a skill for proactive architectural friction discovery and refactoring planning.
Problem / Motivation
Forge has no skill for identifying architectural problems before they become bugs. Current workflow is reactive: issues are created when problems are noticed. A proactive skill would explore the codebase organically, surface friction, and propose targeted refactors.
Inspired by mattpocock/skills improve-codebase-architecture, which combines organic exploration with structured dependency analysis.
Proposed Solution
Create
forge-improve-architecturewith these core concepts:Organic Codebase Exploration
Dependency Categories Framework
Classify each friction point by dependency type, which determines the testing strategy:
Parallel Sub-Agent Design Exploration
"Replace, Don't Layer" Testing
Pipeline Fit
Standalone skill invoked ad-hoc — not part of the linear workflow. Can feed into
forge-create-issuewhen a refactor is identified.Acceptance Criteria