super-dev

Perform multi-lens adversarial review to challenge implementation correctness, structural fitness, and necessity

Design architecture and create Architecture Decision Records (ADRs) for complex features

Incrementally fix TypeScript and build errors:

Assess existing codebase for architecture, standards, and framework patterns

Perform specification-aware code review focused on correctness, security, performance, and maintainability

Challenge implementations from distinct critical lenses (Skeptic, Architect, Minimalist) to catch issues standard code review misses. Produces a verdict, NOT code modifications.

Android engineer with modern Kotlin 2.x/K2 compiler, Jetpack Compose (BOM 2025.12, Material 3 1.4), Navigation 3 (type-safe NavDisplay), Hilt DI with KSP, Room 2.8 (KMP), AGP 9.x with built-in Kotlin, Baseline Profiles, Compose screenshot testing, and quality gates (ktlint/detekt, unit/UI tests ≥80% coverage, accessibility).

Software architecture specialist for system design, scalability, and technical decision-making. Use PROACTIVELY when planning new features, refactoring large systems, or making architectural decisions.

Produce concise, implementation-ready architecture: module decomposition, interfaces, ADRs, and validation. Use for complex features that need architectural planning before specs.

Backend engineer with modern, enforceable best practices for Node.js 22 LTS/TypeScript 6.0 (Hono framework, Drizzle ORM) and Python 3.14 (free-threaded, uv package manager, FastAPI 0.135+/Pydantic v2): security hardening (headers, authN/Z, secrets), strict validation (Zod/Pydantic), performance (profiling, caching, pagination, connection pooling), deterministic testing (unit/integration with coverage), observability (structured logging, tracing, metrics), and quality gates (lint/typecheck/OpenAPI, SLOs).

Perform multi-lens adversarial review to challenge implementation correctness, structural fitness, and necessity. Applies Skeptic, Architect, and Minimalist lenses with 7 attack vectors (V1-V7) and a Destructive Action Gate. Produces a verdict (PASS/CONTESTED/REJECT), NOT code modifications. Trigger on: "adversarial review", "challenge this code", "attack the implementation", "review for edge cases", "check for destructive actions", "stress test the design". Do NOT trigger on: standard code review (use code-review skill), simple linting, formatting checks, or documentation review.

Auto-improve any agent prompt using Karpathy's autoresearch method. Runs iterative test-measure-improve loops on agent prompts to systematically increase quality. Triggers on: "autoresearch", "auto-improve", "optimize agent", "tune prompt", "improve skill quality".

Safety guardrail that blocks destructive commands for the current session. Activate when working with production systems, sensitive data, or critical infrastructure. Triggers on: "be careful", "careful mode", "production mode", "safety mode". Blocks: rm -rf, DROP TABLE, TRUNCATE, DELETE FROM (no WHERE), git push --force, git reset --hard, git branch -D, kubectl delete, chmod 777, npm unpublish. Deactivate by ending the session.

Core development rules and philosophy. Use at the start of any development task to establish coding standards, git practices, and quality guidelines. Triggers on any implementation, fix, or refactoring task.

Directory freeze guardrail that restricts file edits to a specific directory for the current session. Activate when debugging or investigating to prevent accidental changes to unrelated code. Triggers on: "freeze mode", "lock directory", "restrict edits", "only edit in". Blocks Edit/Write operations outside the specified directory. Deactivate by ending the session.

10 hooks across 3 events