SShortSingh.
Back to feed

Tool traces are not an AI agent audit trail

0
·2 views

An LLM observability log can show the prompt, the tool name, and the latency. Useful for debugging. Not enough for an auditor. When someone asks "what did this agent do, and why was it allowed?", you need an authorization envelope at the enforcement boundary: Agent identity plus the human principal or delegation chain Tool + resource + action Policy ID/version and allow/deny with a reason HITL approval when the policy required one Correlation IDs that join the decision to the side effect Logging only after the tool runs misses the control point. Prefer the PDP or MCP gateway that decides befor

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Espacenet Ops: A Scalable Global Patent Search Model

A scalable Espacenet ops framework runs on four sequential stages: Retrieve, Align, Invalidate-check, and Ledger. It is measured by auditable recall per jurisdiction per dollar, not by the number of databases a tool claims to touch. One variable separates a defensible search operation from an expensive one: time-to-defensible-output, not feature count. Everything below maps that principle to concrete workflow architecture, quantitative scoring, failure-mode reconciliation, and a procurement-grade comparison matrix. This is written for people who already run multi-jurisdiction searches: Head of

0
ProgrammingDEV Community ·

Building a Dynamic Strategy Pattern in C# with Customer Tiers and Plugin Architecture

The Strategy Pattern is a staple of object-oriented design. It lets you encapsulate a family of algorithms and make them interchangeable at runtime. However, traditional implementations often suffer from heavy interface hierarchies and rigid factory logic. In modern C#, language features like switch expressions, primary constructors, collection expressions, and collectible AssemblyLoadContext allow us to build clean, extensible strategy pipelines with zero unnecessary boilerplate. In this article, we'll walk through building a flexible shipping rate calculator that supports: Multiple Carrier S

0
ProgrammingDEV Community ·

A listing id that ignores the query string, and the trailing slash that would have made every room new

Notifio is a desktop app that watches rental search result pages and tells you the moment a new listing appears on one. The whole product is a diff. It reads a search page, compares what is on it with what was on it last time, and the difference is the alert. Which means the only interesting question in the codebase is what counts as "the same listing". Get that wrong in one direction and the user gets an email about a room they already replied to.