🇬🇧 English · 🇪🇸 Español
An agentic development harness for Claude Code. One command runs the entire feature pipeline — it collaborates with you on the design, then builds autonomously: plan, execute, double-review, and verify. No per-step babysitting, no choosing which model or skill to use.
/hydraia:feature add rate limiting to the public REST API
You stay on Opus; Hydraia decides the rest — how much ceremony the change needs
(three auto-selected levels), when to drop to Sonnet for execution, which reviewers
to run, which security gates to enforce — and every run ends DONE,
DONE_WITH_FOLLOWUPS, or BLOCKED with evidence, never in an endless fix loop.
Raw Claude Code is powerful but hands-on: you remember to plan, to threat-model, to review, to test — every step manual, every step skippable under pressure. Hydraia applies the discipline you'd use on your best day, on every run. One command runs a fixed pipeline — think → design + threat model → plan → execute → double review → verify — with every model and reviewer decision already made, and a security gate that is always on and never configurable.
| Phase | What happens | Model |
|---|---|---|
| Design | Brainstorm → spec + threat model (scaled by level) | Opus |
| Plan | Plan-as-contract (Files + Verify per task), then frozen | Opus |
| Execute | Fresh sub-agent per task, scope-gated | Sonnet (Haiku for mechanical) |
| Review | Judge + diff-scoped reviewers + security gate → verified triage | Opus judges, Sonnet reviewers |
| Verify | Baseline-aware tests, spec check, secrets/deps scan — bounded | Opus |
Prerequisites (install before Hydraia): Claude Code, git, Node.js ≥18, and
Python 3.8+. codegraph and markitdown are installed for you by
/hydraia:doctor. Platform: macOS/Linux (Windows via WSL).
Add the marketplace and install the plugin — inside the Claude Code CLI or your terminal:
claude plugin marketplace add jdanigo/hydraia
claude plugin install hydraiaThen run /hydraia:doctor once to validate and install external deps. That's it —
every skill and agent ships inside the plugin.
| Command | What it does |
|---|---|
/hydraia:feature <desc> |
Full pipeline: context → design → plan → build → double review + security → verify |
/hydraia:plan <desc> |
Design + threat model + detailed plan, then stop (nothing executed) |
/hydraia:agile <idea> |
Decompose an epic into stories/tasks and deliver it autonomously, stage by stage |
/hydraia:review [focus] |
Double review + security gate on the current branch |
/hydraia:resume |
Continue an interrupted run from the last incomplete phase |
Build & plan
| Command | What it does |
|---|---|
/hydraia:feature <desc> |
Full pipeline end to end: context → think → design + threat model → plan → execute → double review + security gate → verify |
/hydraia:plan <desc> |
Context + design + threat model + detailed plan (with self-review), then stops. Nothing is executed |
/hydraia:agile <idea> |
Multi-stage autonomous delivery: decomposes an epic into stories → tasks and executes the tree by dependency, gated by autonomy tier |
/hydraia:story <story> |
Product-owner analysis of a user story (INVEST, acceptance criteria) → spec → QA cases + traceability matrix → frozen plan, then stops |
/hydraia:architect <idea> |
Greenfield: guided elicitation → architecture options → confirmed stack → API contract → ADRs → full build pipeline |
Diagnose & specialize
| Command | What it does |
|---|---|
/hydraia:perf <symptom> |
Measurement-first performance run: baseline → profile-driven diagnosis → numeric target → implement → re-measure |
/hydraia:db <symptom> |
DB bottleneck run: engine detection, read-only evidence (EXPLAIN, stats, locks), expand-contract migrations |
/hydraia:e2e [focus] |
Generate + run a Playwright critical-flow E2E suite (auto-detected framework) |
/hydraia:devops <request> |
Author CI/CD, Docker, or IaC — deploy and secrets steps flagged for human approval |
/hydraia:observability <request> |
Instrument logs / metrics / traces / alerts — OTel-first, never logs secrets or PII |
Review, understand & docs
| Command | What it does |
|---|---|
/hydraia:review [focus] |
Double code review + security gate on the current branch (code already exists) |
/hydraia:graph <query> |
Query the code graph — call sites, blast radius — without running the pipeline |
/hydraia:explainme <focus> |
Interactive HTML map of a codebase, subsystem, or PR (onboarding in 10 min) |
/hydraia:docs [focus] |
Sync README, API docs, CHANGELOG, and ADR index with the code — reports drift |
Utilities
| Command | What it does |
|---|---|
/hydraia:resume [run] |
Continue an interrupted pipeline from the last incomplete phase |
/hydraia:doctor |
Validate, install, and update external deps (codegraph, markitdown), with consent |
/hydraia:dashboard [port] |
Launch a local (127.0.0.1) web dashboard: plugin status, usage telemetry, editable run modes |
Ship a feature, end to end
/hydraia:feature add rate limiting to the public REST API — 100 req/min per key
Planned, built, reviewed twice, security-gated, and verified — one command.
Lock the approach before writing code
/hydraia:plan migrate the session cache from in-memory to Redis
You get a spec + threat model + frozen plan and nothing is executed. Run
/hydraia:feature when you're happy.
Review a branch you didn't build with Hydraia
/hydraia:review the auth module
Whole-branch + diff-scoped review with the mandatory security gate.
Understand blast radius before touching anything
/hydraia:graph what calls parseConfig and what breaks if I change its signature
Turn a codebase into an onboarding map
/hydraia:explainme the payments subsystem
Questions, ideas, showcases, and release news live in the Hydraia Discord — come say hi:
- CHANGELOG — full release history
- CONTRIBUTING — repo structure + how to add a skill
- License: MIT (Hydraia's own code); upstream skills/agents attributed in NOTICE
- Individual skills are installable into other agents via
npx skills(skills/<name>/SKILL.mdlayout) - A Codex CLI port is available as an experimental preview (
codex/)