rokitoh is the individual author and maintainer behind Red Orbita, a Spanish-language technical publication and open-source security projects. The work connects cloud and platform engineering with practical controls for AI-agent infrastructure.[1]
Start with Bulwark Gateway – fail-closed security proxy for LLM agents →Links
Spanish technical publication and author biographyAuthor and maintainer · Red Orbita
Background
The owned biography describes more than thirteen years of Linux, Unix and Windows systems work and more than five years building and operating Azure infrastructure. It also records experience deploying SIEM systems, performing security testing and hardening production services. Red Orbita turns that operational background into Spanish technical articles and maintained open-source tools.[1]
Contributions
Enforce policy between an application and its agents
Bulwark Gateway places authentication, input checks, threat-intelligence matching, tool permissions, output filtering and rate limits in a proxy before an agent backend. Tenant and agent identifiers select the relevant policies. Streaming tool calls are buffered for validation before reaching the client, and policy failures block the request by default. The documented hot path uses deterministic patterns and validation, with optional machine-learning scanners. Its scope distinguishes these controls from a general web firewall or a guarantee that every malicious instruction will be recognized.[3]
Check a tool call before it executes
The OpenCode Security Agent plugin hooks tool.execute.before and checks proposed calls against local indicators, sensitive paths and command patterns. It can block a call before execution, while a separate static scanner examines skills and integrations before installation. The runtime checks and scanner address different points in the workflow: what an installed agent is about to do and what an extension’s files appear to contain.[5]
Tools, writing & talks
Writing, research & talks
Start here
A self-hosted policy proxy for agent requests, tool permissions and output filtering.
Begin with the architecture and published scope. Development uses Python, Docker and Redis; production deployment adds Kubernetes and explicit tenant/agent policies.
Author and maintainer, confirmed by owned biography[1][3]Original source
Tools
- OpenCode Security Agent
github.comAn OpenCode pre-execution policy plugin and static scanner for skills and integrations.
Author and maintainer[5]