Second Brain OpenCode is a minimal Markdown kit for structuring work around projects, source documents and summaries. OpenCode serves as the single agent; the Markdown files carry the durable memory, so that resuming work does not depend on a previous conversation. Status: usable prototype, published on GitHub since October 4, 2026, whose full functional acceptance testing remains to be redone on OpenCode 2.0.22.

View the GitHub repository →

🎯 Value for the reader
  • A project sheet is the authority for tasks, decisions, log and resume point
  • Original sources stay separate from summaries and are treated as data, not as instructions
  • Two OpenCode rituals: /ouvrir (read-only opening) and /clore (closing that first submits changes for human validation)
  • A fictional example vault lets you walk through how it works without any real personal or professional data

The project provides no vector database, no MCP, no plugin and no permanent automation: these extensions are deferred until a real need is observed.

🧭 Sovereignty and governance

The kit is designed to work with OpenCode and a model reachable through an OpenAI-compatible API on the local network or on the machine itself. This setup can reduce dependence on an external inference provider for the uses concerned, while keeping the real vault separate from the publishable kit.

The approach is inspired by ISO/IEC 27001:2022: separation of sources and summaries, human control before writing, exclusion of secrets from the versioned configuration, preservation of traceability.

No ISO 27001 compliance or certification is claimed. Authentication, transport and logging of the inference service are the responsibility of that service, not of the kit.
🔐 Security and limits
  • The tracked examples are fictional; real data must stay in a separate private vault
  • The local configuration and the risk register are excluded from Git
  • Conversation sharing is disabled in the intended configuration
  • Web calls, shell, plugins, MCP and sub-agents are excluded from the prototype's intended scope
Explicit limits: OpenCode permissions do not form a technical isolation, reading restraint depends partly on the model, and execution paths may exist outside the documented permissions. These measures do not replace controls on the workstation, the network or the AI service.
📜 License

The MIT license (LICENSE.md) allows use, copying, modification and redistribution of the kit, including commercially, provided the copyright and license notice is kept. It provides no warranty and covers only this kit — not the systems, hardware or services it may be used with.

🔎 Validation and publication status
  • Public repository checked on October 4, 2026: main branch, license detected as MIT, published files matching the kit, its documentation and its fictional pilot
  • /ouvrir: partially validated
  • Closing, resuming, summarizing, resistance to hostile instructions, restoration and latency measurement: not tested
  • Full functional acceptance testing to be redone on OpenCode 2.0.22, in its interactive interface
No performance, production compatibility or security guarantee is claimed. The kit is a context-management and traceability prototype, not a compliance product nor a validated security solution.

Related pages

References & Sources

CategoryResourceURL
Public repositorySecond-Cerveau-OpenCode — GitHub (main branch, checked on 2026-10-04)github.com/TurganBarzaloo/Second-Cerveau-OpenCode
Inspiration frameworkISO/IEC 27001:2022 — inspired approach, no compliance or certification claimed—
LicenseSecond Cerveau OpenCode — MIT (this kit only)—
Fact verificationPublic repository and documentation checked on 2026-10-04—
Content of this pageShared under CC BY-SA 4.0creativecommons.org/licenses/by-sa/4.0