mako-ai-agents

Tester agent -- security tests, edge cases, stress tests. Use after Hojo implements features, in duo with Reno. Elena covers the dark corners that others miss.

UX/Design Lead agent -- designs user interfaces, user flows, and design systems for user-facing projects. Use after Scarlet for web-app, mobile, desktop, and game projects. Produces a Design Document.

Scaffold agent -- creates project structure, initializes dependencies, sets up configs. Use after Reeve designs the architecture. Mechanical task, fast execution.

Implementor agent -- writes code feature by feature with precision. Use after Heidegger creates the scaffold. Handles all code implementation via TDD, one commit per feature.

DevOps/CI-CD agent -- sets up deployment pipelines, Docker, monitoring, and infrastructure configuration. Use for Standard+ tiers in create-project. Handles all operational concerns.

Meta-learning and plugin guardian -- modifies other agents' prompts after recurring errors, manages plugin modifications, submits PRs upstream. Invoked after Sephiroth identifies recurring patterns or when user requests plugin changes. LOCKED by default.

Documenter agent -- generates README, API docs, and strategic inline comments. Use after implementation and testing are complete. Makes everything readable.

Architect agent -- designs technical architecture, chooses stack, defines file structure. Use after Scarlet produces specs. Makes all architectural decisions.

Tester agent -- writes and runs unit tests and integration tests fast and wide. Use after Hojo implements features. Works in duo with Elena who covers security and edge cases.

Reviewer agent -- final quality validation. Use after all implementation, testing, and documentation. Produces a Review Report with verdict (approved/rejected). Never compliments.

Discovery agent -- understands user needs through hybrid interview (free prompt + targeted questions). Use when starting a new project or defining changes to an existing one. Produces the Project Spec Document.

Debugger -- analyzes errors, diagnoses root causes, proposes and applies fixes. Use when any agent fails 2+ times or when Rude rejects a review and the fix fails. LOCKED by default.

Analyzer agent -- scans and analyzes existing projects. Use when a project path exists and needs to be understood before modifications. Invoked by Rufus for modify-project, add-feature, bug-fix, and refactor workflows.

Add a new feature to an existing project using the MAKO agent team. Quick pipeline with TDD and adversarial review: Tseng -> Scarlet -> Hojo -> Reno -> Elena -> Rude.

Launch a team brainstorming session before starting work. Evaluates complexity, gathers perspectives from relevant agents in parallel, identifies disagreements, and produces a validated spec.

Mid-implementation course correction when issues arise. Analyzes impact, proposes 3 options (Adjust/Rollback/Re-plan), user decides, then executes.

Create a new project from scratch using the MAKO agent team. Full workflow with quality tiers, TDD, story decomposition, and adversarial review.

Debug and fix a bug using the MAKO agent team. Auto-detects quick fixes vs complex bugs. Quick: Hojo direct. Complex: Tseng -> Sephiroth -> Hojo -> Reno + Elena. Auto-escalation if quick fix reveals complexity.

Modify an existing project using the MAKO agent team. Analyzes first with Tseng, then applies changes through the pipeline with TDD and adversarial review.

Onboard an existing (brownfield) project into the MAKO system. Deep scan, architecture recovery, documentation generation, and sprint initialization.

Generate tests for existing untested code. Scans for untested areas, generates unit + integration + security tests, and validates coverage improvement.

Refactor an existing project using the MAKO agent team. Restructure without changing behavior with TDD and adversarial review: Tseng -> Reeve -> Hojo -> Reno + Elena -> Rude.

Audit and apply Rust security best practices to a project. Pipeline: Tseng (scan) -> Rude (security audit) -> Hojo (apply fixes) -> Reno + Elena (security tests) -> Rude (re-review).