DHDTech AI Intelligence - Autonomous Agents and Skills for Engineering Teams

Overview

The pack of autonomous agents and skills we run across DHDTech.io client engagements: pull-request review, QA subtask generation, pull-request shepherding, migration auditing and AI slop detection. It ships as a macOS installer that drops thirteen skills and seventeen subagents into the engineer's Claude Code and OpenAI Codex setups, and the desktop console refreshes it as new versions publish.

The Challenge

A house standard that lives in a wiki gets ignored. Review depth, delivery discipline and ticket hygiene varied per engineer and per project, and the same mistakes came back in every repository.

The Solution

Every skill ships twice — a Claude Code surface and an OpenAI Codex twin — behind a rule that neither may exist alone. The pack holds thirteen slash-invocable skills and seventeen subagents, with knowledge bases loaded only by the specialist that needs them. Multi-pass pull-request review triages first, fans out eleven passes, auto-approves only when nothing above a low finding survives, and posts the rest as inline GitHub comments. A sibling skill shepherds a branch end to end: open the pull request, wait for CI, nudge the reviewers, merge, watch the deploy. Others generate Jira QA subtasks from a diff, audit Django migrations, detect AI-generated code smells, and build a design system from a guided survey. Four hooks enforce tool batching and skill routing inside the engineer's own sessions. Every push to main versions and publishes the pack to S3, so the console picks it up on its next hourly refresh.