Research noteSource synthesis
Four security boundaries in an AI coding workflow, read more
A source-based map of where untrusted content, agent permissions, tool calls, and security records meet.
For security architects and engineering leads
Research / Evidence desk
Research on AI coding agent security, with evidence labeled.
How to read a note
- Primary guidance
- The cited standard, specification, or vendor documentation.
- Source synthesis
- Rechazo's summary and analysis of those sources, not a lab result.
- Product statement
- Kept distinct from general guidance and linked to Rechazo's site.
Source synthesis
Research notes
No Rechazo product test or independent exploit result is represented here.