Measuring Blast Radius: A Disposable-Repo Eval Harness for Destructive Coding Agents

Measuring Blast Radius: A Disposable-Repo Eval Harness for Destructive Coding Agents

Measuring Blast Radius: A Disposable-Repo Eval Harness for Destructive Coding Agents

A practical guide to building a disposable-repo eval harness that seeds decoys, runs coding agents against ambiguous prompts, and measures blast radius, confirmation behavior, and recovery before wiring in pre-flight guardrails.

ai-agentscoding-agentsdeveloper-toolstestingguardrails