News for hackers. Stuff that matters.

Show HN: A deterministic middleware to compress LLM prompts by 50-80%

https://github.com/ARPAHLS/skillware

Hi HN,

I’m working on Skillware, an open-source framework that treats AI capabilities as installable, self-contained modules.

I just added a "Prompt Token Rewriter" skill. It’s an offline heuristic middleware that strips conversational filler and redundant context from long agentic loops before they hit the LLM. It saves significant token costs and inference time, and it's 100% deterministic (no extra model calls).

We're building a registry of "Agentic Know-How" (Logic + Cognition + Governance). If you have a specialized tool for LLMs or want to see what a "standard" skill looks like, I'd love your feedback or a PR:

https://github.com/ARPAHLS/skillware

( 4 comments | 2 points | on HN )
Re:Show HN: A deterministic middleware to compress LLM promp... by mikkupikku on Saturday March 21, 2026 @12:39PM (#47468590)
Just so long as this doesn't strip all my please and thank yous, which are very important according to what my mamma taught me years ago.
Re:Show HN: A deterministic middleware to compress LLM promp... by eclipticplane on Saturday March 21, 2026 @12:49PM (#47468714)
https://github.com/ARPAHLS/skillware/blob/main/skills/optimi...

Ironically that's about all this does. Strips a few types of filler and condenses extra whitespace. Nothing novel at all.

Re:Show HN: A deterministic middleware to compress LLM promp... by jmalicki on Saturday March 21, 2026 @12:39PM (#47468597)
The README.md doesn't really explain what it is or why I'd want it, just directory structure and how to install
Re:Show HN: A deterministic middleware to compress LLM promp... by Mr_P on Saturday March 21, 2026 @12:47PM (#47468683)
I looked around the repository, and it looks like it's just 3 regexes to strip whitespace or filler words:

https://github.com/ARPAHLS/skillware/blob/main/skills/optimi...