Urgent.News

What's breaking now, across thousands of outlets.

AI

Declare `tools` Only When Narrowing an APX Agent

Declare tools Only When Narrowing an APX Agent An agent definition should describe a durable role without freezing a temporary snapshot of the runtime around it. That is why APX treats an omitted tools field differently from a declared list. The rule is small: omit tools when the agent should inherit APX's broad project-agent default. Declare a non-empty list only when you intentionally want a…

APX treats a missing tools field differently from a declared list of tools in an agent definition. The omission preserves the agent's broad default tool allowlist from the runtime, while a declared list acts as a narrowed allowlist. This distinction connects the portable context layer (APC) and the daily-use runtime and tooling layer (APX).

Omitting tools avoids coupling the agent to specific runtime capabilities, allowing new runtime capabilities to become available without modifying the agent definition. Declaration is useful for agents with genuinely constrained jobs, such as read-only reviewers or documentation inspectors. However, a narrow list should not stay narrow when APX adds capabilities; copying every currently available tool creates maintenance cost without adding meaningful boundaries.

Capability is not determined by prompt size, as APX separates these concerns. The allowlist answers which tools the runtime may permit, while lightweight channels can start with a small base set and discover additional tools when needed. When reviewing .apc/agents/*.md, ask why each tools list exists. If the answer names a boundary, keep it.

If the answer is that the tools were available when the agent was created, remove the field and let APX resolve the default. Runtime-specific overrides can be stricter for a particular run, including running with no tools. The APC file should state constraints that should survive across machines, while a declared list says this boundary must not.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

Building a $5.70/Month Autonomous AI Coding Agent in the Cloud Using Open-Source Terminal Agent Management

Originally published on tamiz.pro . Introduction Autonomous AI coding agents promise to reduce developer friction by running continuously in the cloud, writing, testing, and refactoring code without…

  • Autonomous AI coding agent costs $5.70/month using open-source tools
  • Ubuntu virtual machine on DigitalOcean or Vultr provides infrastructure
  • Agent-deck and orca enable autonomous coding, testing, and refactoring

How to Run an AI Model on Almost Any Hardware: A Practical Guide to Tiny Local LLMs

This month a developer ran a working language model on a Sony PSP-2000. That is a 333 MHz MIPS handheld from 2007 with 64 MB of RAM, and it is streaming English text onto its LCD at one to two tokens…

  • Developer ran language model on 2007 Sony PSP handheld device with 333 MHz MIPS processor
  • Karpathy's stories15M model int8-quantized to 17 MB, runs in 1,100 lines of C code
  • Guide provides hardware-first approach for running tiny local LLMs on various hardware

More from Sunday 6 September →