AgentAmb
An instruction-ambiguity expander that takes any vague agent command and ranks every plausible interpretation by blast radius.
AgentAmb takes an ambiguous instruction — like "clean up the cost explorer" — and expands it into all the plausible things an AI agent might actually do, each ranked by blast radius from SAFE (read-only) to CATASTROPHIC (irreversible). It is pre-loaded with the Kiro/AWS June 2026 incident, where five words triggered a deletion command that took billing offline for 13 hours, and a "Make it precise" button reveals a safe rewrite for every scenario. Pure single-file HTML and JavaScript with zero dependencies; an optional live mode lets you paste your own Gemini key to expand custom instructions entirely in-browser.
The build log
Follow the work.
New essays and prototypes, in your feed reader. Email subscriptions are paused while I connect a working delivery service. No email is collected here.
Follow via RSS