PUBLIC PROJECTS BY BLUETHROAT LABS
Free tools and practical research for safer work in Web3.
PUBLIC RESEARCH PROJECT
Understand how trusted execution environments fail.
Inspect an agent skill before granting access.
Redact sensitive image content before sharing.
COMING SOON
View untrusted PDFs through isolated rendering.
Assess unfamiliar code before running it.
A free, public security handbook for teams building with trusted execution environments.
Explore how TEEs fail in practice, the assumptions behind attestation, and the engineering choices that shape the security of TEE-based Web3 systems. The handbook brings together practical research, failure modes, and guidance for developers and security researchers.
Assess an AI agent skill before giving it access to your environment.
Inspect a skill for suspicious instructions and risky behavior before installing it. Use the assessment to understand identified risks and what remains uncertain.
LIMIT
An assessment does not guarantee that a skill is safe in every environment.
Redact sensitive parts of an image before sharing it.
Use secure masking to replace covered pixels in the exported image. Cosmetic blur is available for appearance. An offline workflow lets you disconnect before importing a sensitive image.
Cosmetic blur is not secure redaction. Uncovered content and mask geometry remain visible.
A project for viewing untrusted PDFs through isolated rendering.
BluePaper is being developed to help people inspect documents from unfamiliar senders through an isolated rendering workflow.
Rendering isolation does not establish that a document's content, links, or sender can be trusted.
Assess an unfamiliar GitHub repository before running its code.
BlueGate is being developed to help people inspect unfamiliar repositories for suspicious behavior and understand the risks of running them.
An assessment is not a guarantee that code is safe.