agentlint

Detect AI/agentic-specific anti-patterns that traditional linters miss. Analyzes tool/agent boundary violations, prompt debt, context window issues, testing patterns, and more. Returns scored findings with remediation guidance.

Detect unintegrated features before they become dead code. Catches the "Ink component pattern" - code that exists but isn't wired.

Guides test type selection for agentic code. Use when writing tests for LLM interactions, tool implementations, or behavioral quality. Covers unit tests, VCR integration tests, and TruLens evals per ADR-0011/0012.

Review test quality before PR - semantic analysis of test design for agentic code. Checks test type appropriateness (unit vs VCR vs evals), ADR-0011/0012 compliance, and Constitution alignment. Run as Level 2 review after epic implementation.

A test skill for validation

Research architecture decisions and create Architecture Decision Records (ADRs). Use when facing complex technical choices requiring deep research, options analysis, and formal documentation. Treats user as CTO for collaborative decision-making. Supports evolutionary architecture - can challenge prior decisions when evidence warrants.

Analyses Arc42 architecture documentation and decomposes it into well-sized Epics suitable for Speckit implementation. Use when breaking down architecture into implementable work, creating epic catalogues from design docs, preparing Arc42 for agile delivery, or when asked to plan epics from architecture.

Decompose high-level design documentation into detailed Arc42 architecture documentation. Use when the user asks to create architecture documentation, generate Arc42 documents, analyze a codebase for architecture patterns, or convert existing design docs to Arc42 format. This skill analyzes repositories, identifies architectural components, validates context and assumptions through user questions, and produces comprehensive Arc42-compliant documentation.