Orchestrating Linear + GitHub with AI Agents
Navigate: arrow keys · Back to site
| What we KNOW | What we DO | What gets ENFORCED |
|---|---|---|
| Commit conventions | Sometimes | Almost never |
| No secrets in code | After the incident | Sporadically |
| Tests before merge | On new projects | Until there's pressure |
| Traceable issues | On "important" PRs | When there's an audit |
| Definition of Done | In retros | Never mechanically |
"The agent doesn't just write code — it self-enforces the rules your team already knows but can't consistently follow."
CLAUDE.md = The operations manual
Skills = The menu recipes
Harness = Quality control
Hooks = Automatic sensors (oven alarm, timer)
The chef (agent) follows the recipes, quality control verifies,
and the sensors fire on their own.
git commit — API key in code
git commit — clean code, key in .env
deleteTask
/close-issue DEMO-1
npm test → PASS
gh run list → PASS
Linear API → PASS
"If your tools don't give you control, build your own — it's simpler than you think."
The entire enforcement system fits in <900 lines.
No frameworks. No dependencies. Full control.
Fork it. Star it. Adapt it.
github.com/felirangelp/harness-driven-dev
Setup: ~2 hours · Harness: <900 lines · ROI: immediate
Made by Felipe Rangel