Build. Test. Rethink.
How useful agents fail—and which defenses preserve their usefulness.
Independent research for engineers giving AI agents access to code, private data and tools. Documented failures, practical defenses, and the evidence behind deployment decisions.
Start here
Access is only the beginning.
Follow the failure
Your agent can have permission to do the wrong thing
Which boundary separates valid access from authorized disclosure? A reported GitHub MCP case and four layers of authority.
Read the analysis →
Examine the defense
What does your agent's sandbox actually contain?
Read access, write access and remote authority are different boundaries. A constructive review of what a sandbox can—and cannot—establish.
Read the analysis →
Three questions we follow
Follow the failure
Which permitted capabilities let a reported attack reach its objective? We trace the task, authority and path to the outcome.
Examine the defense
Where is the control enforced? We compare what it prevents, what useful work it preserves, and what it costs.
Challenge the assurance
What does the evidence actually establish? We separate observed behavior from coverage claims and deployment judgments.
Our evidence standard
A useful conclusion comes with its limits.
We link primary sources, distinguish reported findings from our own analysis, and explain the conditions under which a defense helps. A proposed experiment is never presented as a result.