Jengo Agent

Inspiration

Running Jengo AI , a generative AI community for creatives, coders, founders, and dreamers, means constantly context-switching between two very different modes of work: creative output (scripts, copy, content) and execution (tasks, issues, sprints). Every day I'd lose momentum switching between tools. I built Jengo Agent to collapse that gap into a single prompt.

What it does

Jengo Agent is an AI agent built on the GitLab Duo Agent Platform with two core modes:

Creative Mode - generates ad scripts, content briefs, taglines, and campaign copy, tuned for TikTok, Substack, Twitter, and Discord workflows
Forge Mode - produces GitLab-ready issues, user stories, acceptance criteria, sprint plans, and SEO briefs A smart Auto router detects intent from your prompt and routes to the right mode, no configuration needed. Every response ends with a Next Action so you always know the immediate next step.

How we built it

The agent is built on the GitLab Duo Agent Platform with a YAML trigger config that fires on /jengo slash commands in MR comments and issue threads. The intelligence layer uses Claude (Anthropic) via the Messages API, with a dual-mode system prompt that enforces structured, deliverable-focused output. The frontend is a React artifact with persistent conversation history passed on every API call for full multi-turn context.

Challenges we ran into

The hardest part was prompt engineering the dual-mode boundary, making the agent feel opinionated and specialized in each mode without being rigid when a prompt spans both. The solution was a single system prompt with mode-lock overrides, plus requiring the agent to self-tag every response so the user always knows which mode fired.

Accomplishments that we're proud of

a single system prompt with mode-lock overrides

What we learned

Building on the GitLab Duo Agent Platform reinforced how much value lives in the trigger layer, the right agent at the right moment in the dev workflow (MR review, issue triage, sprint planning) creates compounding leverage that a standalone chat interface can't match.

What's next for Jengo Agent

GitLab webhook integration to auto-create issues directly from Forge Mode output
Scheduled Substack brief generation triggered by a GitLab pipeline cron
JengaForge integration for end-to-end idea-to-shipped-feature tracking