CSBP intentionally splits its responsibilities across three files so that system definition, runtime application, and practice formation can never collapse into each other. Each file has a fixed role it owns exclusively. None of the three borrows from the others at runtime, and none can be modified to take on a role that belongs to another. This separation is not a structural preference — it is the mechanism that keeps the system stable.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/xxyoudeadpunkxx/csbp-codex-shared-best-practice/llms.txt
Use this file to discover all available pages before exploring further.
Why three files?
Three jobs must stay permanently separate inside CSBP:- Define the system. Explain what CSBP is, how it is loaded, and what authority it holds. This job belongs to the bootstrap layer.
- Apply promoted practices. Use only already-approved runtime instructions when they are relevant to current work. This job belongs to the runtime layer.
- Form or revise practices. Evaluate whether a recurring pattern deserves promotion into the runtime layer. This job belongs to the formation layer.
File overview
| File | Role | Read timing | Mutable at runtime |
|---|---|---|---|
CSBP-entry-point.txt | Bootstrap and routing layer. Defines system identity, authority order, load order, mode routing, and write restrictions. | First, every session. | No |
shared-best-practice.txt | Runtime layer. Stores already-promoted practices that may guide live work when relevant. | After entry point. | Yes, but only through compiler path and operator approval. |
shared-best-practice-compiler.txt | Formation layer. Governs evaluation, normalization, promotion, revision, deprecation, removal, and rejection of candidate practices. | Only during practice formation work. | No during normal runtime. |
CSBP-entry-point.txt
CSBP-entry-point.txt is the bootstrap and routing layer. It is the structural anchor of the entire system. Its job is to define what CSBP is, establish the authority order, set the read order for all three files, route Codex into the correct usage mode, and enforce write restrictions. It defines system identity so that no other file needs to — and so no other file can override it.
The entry point covers:
- System identity and object
- Authority chain
- Layer roles for all three files
- Usage modes and when each activates
- Entry decision and entry rule
- Read order
- System definition lock
- Runtime, compiler, empty-state, conflict, and write rules
CSBP-entry-point.txt is the first thing read every session. Reading it activates CSBP for the current session and determines which mode loads next.shared-best-practice.txt
shared-best-practice.txt is the runtime store for already-promoted practices. It is read after the entry point during every session and is the only file that changes during normal CSBP use — and only with operator approval after a practice has passed through the compiler.
Active practices in this file may guide work when their applies_when condition matches the current operation. Deprecated practices remain in the file for review history but are excluded from runtime use.
Each promoted practice uses this fixed block shape:
shared-best-practice-compiler.txt
shared-best-practice-compiler.txt is the practice formation and evaluation layer. It governs how candidate practices are evaluated, normalized, promoted, revised, deprecated, removed, or rejected. It is the decision layer that determines whether a recurring pattern is strong enough, stable enough, and well-formed enough to enter the runtime store.
The compiler is not loaded during normal runtime work. It is only activated when the operator explicitly starts practice formation or maintenance work. This keeps the layer light during ordinary sessions — the compiler’s evaluation logic does not bleed into active runtime guidance.
Like the entry point, the compiler file is fixed. It is never modified or extended at runtime. It is a system-definition layer that shapes what enters the runtime store, not a store itself.