AI Engineer · October 12, 2026

Agent workflows: review step permissions and evaluate output quality separately

video Agent Security AI Engineering Model Evaluation

Agent workflows: review step permissions and evaluate output quality separately video thumbnail

Why it matters

Flywire’s engineers describe Orbit workflows that combine deterministic steps with restricted model calls. Reviewed YAML declares tool and network access; timeouts and locks bound execution and duplicate work. Historical traces feed evaluations of individual model steps. The speakers identify incomplete sandboxing and online evaluations as remaining work. Their execution-success figures do not establish answer correctness or prove the controls contain an attack.

My takeaway: For a read-only review step, test attempted file edits, disallowed tools, and unauthorized network access. Replay duplicate triggers and malformed structured outputs. Evaluate saved cases for correctness as well as completion, and require fresh checks when permissions, prompts, models, or workflow code change.

Keep exploring

More curated notes connected through Agent Security and AI Engineering.

OpenAI News · framework

Pacing model development in an era of cyber-critical capabilities

OpenAI says preliminary evidence that Astra may meet its Critical cybersecurity threshold led it to pause frontier reinforcement-learning work for two weeks and keep its largest planned run on hold. New safeguards include stronger workload and network isolation, continuous boundary testing, token-level monitoring that escalates suspicious tool activity, and broader alignment checks for deception, reward hacking, and unauthorized access.

OWASP GenAI Security Project · guide

OWASP Top 10 for Agentic Applications for 2026

OWASP's community guide organizes agentic-system risk into ten categories, including goal hijacking, tool misuse, identity and privilege abuse, memory poisoning, insecure inter-agent communication, cascading failures, and rogue-agent behavior. It provides a shared taxonomy and mitigation starting point rather than a certification checklist or evidence that a deployed system is secure.

OpenAI News · framework

A blueprint for democratic governance of frontier AI

OpenAI proposes a three-part U.S. frontier-AI governance model: harmonize emerging state safety laws into a federal baseline, strengthen CAISI as an evaluation and standards institution, and coordinate a broader resilience program. Proposed controls include severe-risk evaluations, transparency reports, independent audits, safety-incident reporting, model-weight security, whistleblower protection, and periodic technical assessments.