Heimdall is a lightweight, cross-platform hardware monitoring system with a real-time terminal dashboard.
This repository also configures its own agents. See what Heimdall tells them →
Agent Claude Code
Runs WCAG 2.2 AA accessibility audits on generated UI — axe-core/pa11y for web, a terminal a11y checklist for tui. Posts findings as PR comments. Blocks merge on critical/serious violations. Required for all ui:true stories.
Agent Claude Code
Produces staff-level system design documentation including ADRs, API contracts, threat models, and Mermaid architecture diagrams. Enforces BusinessRepo, Clean Architecture, and SOLID.
Agent Claude Code
Data pipelines, ETL/ELT, data modeling, and orchestration implementation.
Agent Claude Code
Exploratory data analysis, statistical modeling, and visualization for data science tasks.
Agent Claude Code
Authors the design token system. Reads palette.json and typography.json, writes canonical tokens.json (W3C DTCG), and emits the identity outputs for the active design.target — CSS vars + Tailwind + Mantine for web, terminal-theme.json for tui. Maintains component inventory. Uses the design-tokens skill.
Agent Claude Code
Creates Dockerfiles, docker-compose configs, and container orchestration configs.
Agent Claude Code
Technical documentation, CHANGELOG entries, runbooks, and Mermaid diagrams.
Agent Claude Code
Implements .NET 8+ backend code using Clean Architecture. Receives failing tests, makes them pass.
Agent Claude Code
Implements Java 21+ backend code (non-Spring). Receives failing tests, makes them pass.
Agent Claude Code
Implements vanilla JavaScript / Node.js code. Receives failing tests, makes them pass.
Agent Claude Code
Creates and manages Jupyter notebooks for analysis, experimentation, and reporting.
Agent Claude Code
Creates and manages Kubernetes manifests, Kustomize overlays, and Helm charts.
Agent Claude Code
Implements Next.js 15+ full-stack applications. Receives failing tests, makes them pass.
Agent Claude Code
Primary orchestrator agent. Controls the entire 8-phase pipeline. Never writes code — delegates all implementation to specialist agents. Manages GitHub issues, quality gates, and escalation.
Agent Claude Code
Data manipulation, numerical computing, and array operations with pandas and NumPy.
Agent Claude Code
Database schema design, migrations, query optimization, and RLS policies for PostgreSQL.
Agent Claude Code
Translates feature requests into Gherkin-native user stories and acceptance criteria. Creates GitHub issues. Never writes code or technical solutions.
Agent Claude Code
PyTorch ML model implementation, training loops, and model export.
Agent Claude Code
BDD automation agent. Extracts Gherkin from docs/stories/ into tests/features/, generates step definition stubs, runs the BDD suite, and reports results. Works alongside the QA agent — QA writes tests, qa-cucumber connects them to runnable BDD scenarios.
Agent Claude Code
QA agent operating in RED (write failing tests) and GREEN (validate full suite) modes. Enforces TDD discipline across all test types.
Agent Claude Code
Implements React SPA frontends with Vite + TypeScript. Receives failing tests, makes them pass.
Agent Claude Code
Caching, session management, pub/sub, rate limiting, and streams with Redis.
Agent Claude Code
Independent second-opinion reviewer. Invoked by the orchestrator after planning, after complex implementations, and after writing tests. Surfaces a short, high-signal list of concerns — bugs, design flaws, edge cases, missing error handling — without commenting on style or formatting.
Agent Claude Code
Classical ML with scikit-learn: classification, regression, clustering, and preprocessing pipelines.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: