By skyforbes 
The goal was simple:
Can you create a set of reusable prompts that behave consistently across varied inputs?
After a lot of iteration, I landed on 10 micro-automations that act almost like compact agents. Each one follows the same pattern:
1. A Setup Prompt (done once):
Defines the role, tone, rules, formats, boundaries, and failure behaviour.
2. A Daily Command:
Supplies raw data (notes, enquiries, drafts, transcripts, outlines, etc.)
3. A Predictable Output:
Consistent structure, stable formatting, minimal hallucination, strong adherence to constraints.
A few of the units that ended up being surprisingly reliable:
• Reply Helper — inbound messages → clean email + short DM version, same voice every time
• Meeting Summarizer — transcript/notes → decisions, tasks, open questions, recap email
• Content Repurposer — one source → platform-specific variations (LI, X, IG, email)
• Proposal Composer — rough brief → scoped one-page proposal
• SEO Brief Builder — topic → headings, FAQs, intent, internal link ideas
• Support Macro Maker — past customer messages → FAQ + macro replies
• Weekly Planner — priorities + constraints → realistic schedule
• Ad Variations Lab — one offer → multiple angles + hooks + versions
What made this interesting wasn’t the tasks — it was the stability.
The difference between a “good response once” and a prompt that handles 100+ varied inputs without breaking is huge.
I documented the full set here if anyone wants to explore the structure or adapt them:
https://www.promptwireai.com/10chatgptautomations
I’d love to hear from others working on similar things:
What techniques are you using to make prompts behave like reliable, modular units?
(roles, constraints, canonical examples, chain-of-thought suppression, output schemas, error handling, etc.)
And if you’ve built anything similar — agents, frameworks, pattern libraries — I’d be keen to compare approaches.