See which secrets a coding agent can reach
Heldkeys lists the credentials an AI agent can access on a machine or in CI; aws-safebox is a proof-of-concept command guardrail for AWS work.
Source: Project pages
Tools for this
HHeldkeys
Open-source tool that lists the credentials an AI agent can reach on a laptop, sandbox or CI and flags risky agent steps in GitHub workflows.
Aaws-safebox
Proof of concept: local guardrails that allow or block commands an AI agent runs during AWS infrastructure work; the author says it is not a hardened sandbox.
FAQ
Where does this use case come from?
Project pages: https://github.com/hugomn/heldkeys
Is this a guaranteed result?
No. It describes what a builder reported or what vendors say their tools do. Results depend on your setup.
Get new tools and use cases by email
A free email, weekly, every Friday, with newly added AI tools and agent use cases. Unsubscribe anytime. Plus our free AI tools starter kit for small businesses (PDF).