prompt-ctl (@promptctl) header
prompt-ctl (@promptctl) profile picture

CLI for LLM prompt engineering. Structure prompts, save ~67% on Claude, GPT, Groq. One binary, Homebrew. https://nitter.cf/t.co/schbb30Qsh #builtinpublic #opensource #ai

Netherlands
Joined February 2026
Photo by prompt-ctl (@promptctl)Photo by prompt-ctl (@promptctl)Photo by prompt-ctl (@promptctl)Photo by prompt-ctl (@promptctl)Photo by prompt-ctl (@promptctl)Photo by prompt-ctl (@promptctl)Photo by prompt-ctl (@promptctl)Photo by prompt-ctl (@promptctl)Photo by prompt-ctl (@promptctl)
Prompt engineering today feels like backend systems before CI. Most teams: - Copy prompts between docs - Tweak them in production - Have no regression tracking That works- until it doesn’t.
1
1
1
3
58
prompt-ctl retweeted
I built @olegkoval/queryd to catch DB latency before users feel it. What it does: - detects slow queries across Node DB clients - tracks per-request query budgets (query storms) - works with Prisma OR driver-agnostic wrappers 60-second setup in README: github.com/oleg-koval/slow-q…
1
2
3
46
Prompt engineering today feels like backend systems before CI. Most teams: - Copy prompts between docs - Tweak them in production - Have no regression tracking That works- until it doesn’t.
2
14
Example workflow: Modify prompt → v2 Run against v1 baseline CI fails if quality drops Now prompt changes are reviewable, measurable, reversible.
1
7
LLM systems are becoming production systems. Production systems need regression protection. Prompt workflows are at the same maturity transition that CI once was. #devops #promptctl #promptengineering #ai #llms Full write-up: olko.substack.com/p/your-pro…
4
Built RateCard.pro - a marketplace prototype where EU senior contractors publish transparent rate cards and clients pay to unlock direct contact, cutting out recruiters entirely. @bubble kvlolg.bubbleapps.io/version…
1
18
Yes. We support checking prompt quality before send. Every promptctl send run scores the prompt (0–100) and shows Prompt quality: N/100 in the pre-send block. You can also require a minimum score with --min-score=N so the CLI refuses to send when the score is below that
Replying to @promptctl @ffugenw
can it check prompts quality score for desired output
1
11
Score prompts on disk (same scoring logic) promptctl score path/to/prompts promptctl score . --min-score=80
1
6
So: we do support it, and you can test by running promptctl send (with or without --min-score) and optionally promptctl score on a directory. @harshfagarwal
6
promptctl now supports Gemini 3.1 Pro in 0.9.1 release • --thinking-level (low/high) for reasoning control • --media-resolution for vision/multimodal • Multimodal function responses (images + PDFs) CLI for structured prompts, fewer retries. prompt-ctl.com
1
6
promptctl now supports Gemini 3.1 Pro. What's new: • --thinking-level for reasoning control • --media-resolution for vision/multimodal • Multimodal function response support Update your CLI and build smarter agentic workflows. prompt-ctl.com
3
Building promptctl has been a good reminder: Most dev tools fail not because of missing features, but because the first 60 seconds are confusing. So in v0.8 I didn’t add “AI magic.” I added: onboarding wizard model picker prompt rating loop local prompt memory Now the flow is:
2
1
6
40
"promptctl create "Build a SaaS landing page"" → get structured prompt → rate output → retry if needed → save to memory No dashboards. No prompt spreadsheets. Just a CLI workflow. #BuildInPublic #IndieHacker #DevTools #AI
1
1
6
Still early, still rough in places, but the direction is clear: prompts should behave more like code, less like chat history.
1
4