skills.rest
sundarshahi
sundarshahiCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

drydock

Orchestrates 19 specialized agents to build, test, secure, and ship software systems.

1|Updated Jun 24, 2026
Search Tags:#devops#ci/cd#code generation#security audit#multi-agent#orchestration#pipeline
One-click install
npx skills add https://github.com/sundarshahi/drydock --skill drydock-sundarshahi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: drydock
Source: https://github.com/sundarshahi/drydock/tree/main/skills/drydock
Command: npx skills add https://github.com/sundarshahi/drydock --skill drydock-sundarshahi

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Turning a one-line idea into a production-ready system normally requires coordinating product management, architecture, backend, frontend, QA, security, DevOps, and documentation work. This Skill orchestrates that entire workflow inside Claude Code, routing requests to 19 specialized agents so a plain-English request produces architecture docs, tested code, security audits, CI/CD pipelines, and launch materials instead of raw code files. ## Core Features & Use Cases - Request classification into 14 execution modes: Full Build, Feature, Harden, Pentest (VAPT), Compliance, Ship, Test, Review, Architect, Design (UX), Document, Explore, Optimize, and Launch (GTM), each routing to only the relevant agents. - Six-phase pipeline with 3 approval gates: DEFINE, BUILD, HARDEN, SHIP, LAUNCH, and SUSTAIN phases pause at BRD, Architecture, and Production Readiness gates, with receipt verification and rework loops. - Continuous security enforcement: a PreToolUse secret-guard hook blocks writing secret-bearing files, and a BUILD-exit gate runs SCA, secret scanning, and SAST before code advances. - Use Case: Describe "Build a SaaS for booking dog walkers with auth, payments, and a dashboard" and the orchestrator bootstraps a workspace, asks your autonomy level, then runs agents that produce the BRD, OpenAPI contracts, tested services, Dockerfiles, Terraform, and go-to-market docs. ## Quick Start Ask the AI to build your product idea in plain English, for example by saying you want to build a SaaS app with authentication, payments, and a dashboard, then pick an autonomy level when prompted.

Frequently Asked Questions about drydock

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a full-stack app from a one-line description?▼

Describe the product in plain English and the orchestrator classifies it as a Full Build, asks your autonomy level, then runs the DEFINE through SUSTAIN pipeline. Nineteen specialized agents produce the BRD, architecture, tested code, security audit, CI/CD, and documentation, pausing at three approval gates.

What is the difference between Full Build and Feature mode?▼

Full Build runs all 19 agents across six phases for greenfield products. Feature mode is scoped to existing code and runs only product management, architecture, backend or frontend, and QA agents for the requested feature.

Can I run only a security audit or pentest on existing code?▼

Yes. Harden mode runs security, QA, and code review agents in parallel against existing code, while Pentest (VAPT) mode runs an 8-phase security test including live DAST. Pentest requires an authorization gate before any active testing begins.

Does it work with an existing brownfield codebase?▼

Yes. The bootstrap step detects existing source files, generates a .drydock.yaml path mapping, and writes a codebase-context file instructing agents to match existing style and never overwrite files without approval. Greenfield projects are created from scratch instead.

How does the pipeline prevent secrets from being committed?▼

A PreToolUse hook hard-blocks writes and commits to secret-bearing paths like .env, .pem, and credentials.json, and scans staged diffs with gitleaks or a regex fallback. A BUILD-exit gate additionally runs SCA, secret scanning, and SAST before code advances.

What are the limitations of the parallel agent execution?▼

Maximum parallelism with worktree isolation requires a clean git state; dirty repos prompt you to auto-commit or run in a shared directory. Remediation loops cap at two cycles before escalating to the user, and gate breaches block shipping unless an override receipt is logged.

Related Skills

Explore More Software Engineering Skills→
ddarkrddarkr
5

oma-brainstorm

Explore design ideas and constraints to document approaches before coding.

Community
Intermediate
diriendirien
19

sre-engineer

Define SLOs, SLIs, error budgets, and Prometheus alerting rules for production services.

Community
Advanced
builders-gardenbuilders-garden
43

siwa-server-side

Verify SIWA signatures and issue ERC-8128 receipts for API routes.

Official
Advanced
QBall-IncQBall-Inc
8

create-subagent

Generate single-purpose Claude Code sub-agents with validation and deployment.

Official
Advanced