Hacker News Reader

jayanaka98on Agent skills should be compiled, not just read

So true. A lot of skills out there already mix methodology and orchestration together. SIGIL’s AG-IR compilation helps extract the methodology and keep that separation clean.

jayanaka98on Agent skills should be compiled, not just read

Ha ha, true. We already had the solution for compliance years ago with LangGraph, byLLM, etc. I think we went with SKILL.md because of the ease of authoring and use. SKILL.md or AGENT.md is genuinely easier to use, with a low barrier to entry for building custom workflows, though not very compliant. SIGIL seeks to address this dilemma by compiling agent skills.

jayanaka98on Show HN: Sigil – compile a SKILL.md the model MUST run, not "might"

Sigil compiles a plain-markdown skill into a program the model runs inside. It's not a runtime that reads skills, it's a compiler: SKILL.md in, standalone agent harness out.

It's written in Jac (https://www.jaseci.org), that's where the graph runtime, the typed "by llm" slots and the concurrency come from. Docs: https://sigilagent.com. Happy to go deep on the IR design.

jayanaka98on Ford AI hiccups push carmaker to rehire ‘gray beard’ inspectors

The reason AI fails in Industry is that SKILL.md or other knowledge-injection methods do not guarantee compliance. AI just thinks it "knows better".