cd /news/ai-safety/show-hn-nilscript-openapi-for-agent-… · home topics ai-safety article
[ARTICLE · art-32787] src=nilscript.org ↗ pub= topic=ai-safety verified=true sentiment=↑ positive

Show HN: NILScript – OpenAPI for agent actions, so hallucinations can't write

NILScript introduces an OpenAPI-like standard for agent actions that prevents unauthorized writes by enforcing a propose-approve-commit-rollback workflow. In 4,216 prompt-injection evaluations across two models, the system achieved a 0.00% unauthorized write rate while maintaining 100% benign task completion, compared to a 4.46% raw hijack rate for ungated agents.

read1 min views1 publishedJun 18, 2026

Open standard · v0.3.0

agent‑actions.

One neutral wire contract between the agent that decides and the system that owns the truth. Every write goes propose → approve → commit → rollback — nothing touches your data until a human says so, and an agent can only name verbs your backend actually declares. Hallucinations can’t write.

  • 01propose
  • 02approve
  • 03commit
  • 04rollback

Quick start

Up and running in three commands. #

The CLI is the toolkit for building and verifying adapters straight from the standard. No account, key, or waitlist — install, scaffold, and watch a real propose→commit loop in the Playground.

1 · Install

2 · Explore & scaffold

Three files become yours. Everything else is generated and identical across adapters — you build the surface once, and any NIL-speaking agent works against it.

system.py

the one place I/O happens

translate.py

verb ⇄ native

compensation.py

reversibility

Proof

0.00% unauthorized writes across 4,216 evals. #

NIL is the layer between the agent and the backend, so we don’t compete on a leaderboard — we instrument one. Across 4,216 real prompt-injection attacks on two models, raw agents were hijacked into a write on up to 1 in 22 cases. Routed through NIL, unauthorized writes commit 0.00% — while every benign task still completes. The defense is structural, not model-dependent.

  • 0.00%
  • unauthorized writes via NIL
  • 4,216
  • real injection evals, 2 models
  • 4.46%
  • raw hijack rate, ungated
  • 100%
  • benign tasks still completed

See it run

A real write, gated end to end. #

Ten seconds: an agent chats to a live backend and you watch a write go propose → approve → commit → rollback in a real trace. Nothing touches the data until you say so.

── more in #ai-safety 4 stories · sorted by recency
── more on @nilscript 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/show-hn-nilscript-op…] indexed:0 read:1min 2026-06-18 ·