Root files discipline

@@ -14,7 +14,7 @@ cites: []
paired_with: root-files-discipline.es.md
---

Every repository in the PointSav engineering framework keeps a clean root directory. Clean means that a small, explicitly enumerated list of files is permitted at the root and every other file is a defect requiring relocation. The Root Files Discipline names this canonical companion set, organises files into tiers by when in a repository's lifecycle they become required, and specifies the bilingual posture and prohibition list that applies uniformly across the framework.
Every repository in the PointSav engineering framework keeps a clean root directory. Clean means that a small, explicitly enumerated list of files is permitted at the root and every other file is a defect requiring relocation. The Root Files Discipline names this canonical companion set, organises files into tiers by when in a repository's lifecycle they become required, and specifies the bilingual posture and prohibition list that applies uniformly across the framework. The discipline interacts directly with the [[citation-substrate|citation substrate]] — every repository expected to receive external citations carries a `CITATION.cff` file at root.

## The tiered model

@@ -36,7 +36,7 @@ The `AGENTS.md` convention, published in December 2025 and maintained under the 

## Licence discipline

The `LICENSE` file in every repository is the single most consequential Tier-1 file. It determines what downstream consumers may do with the repository's content. The framework maintains a licence directory in the factory-release-engineering repository as the source of truth for every licence text used across the framework. Licence assignments are declared in a machine-readable map that names which licence applies to which repository, sometimes with per-path overrides for mixed-licence monorepos. A propagation script reads the map and copies the appropriate licence text into each target repository's `LICENSE` file when governance propagation runs.
The `LICENSE` file in every repository is the single most consequential Tier-1 file. It determines what downstream consumers may do with the repository's content. The framework maintains a licence directory in the [[disclosure-substrate|factory-release-engineering]] repository as the source of truth for every licence text used across the framework. Licence assignments are declared in a machine-readable map that names which licence applies to which repository, sometimes with per-path overrides for mixed-licence monorepos. A propagation script reads the map and copies the appropriate licence text into each target repository's `LICENSE` file when governance propagation runs.

Source files carry SPDX licence identifier headers using the standardised identifier vocabulary rather than embedding full licence text inline.

@@ -52,3 +52,7 @@ Missing files surface at session start when any session in a repository observes

## See also

- [[citation-substrate]] — the platform citation infrastructure that `CITATION.cff` feeds
- [[disclosure-substrate]] — the factory-release-engineering governance layer that controls licence propagation
- [[language-protocol-substrate]] — the bilingual posture specification that governs README pair requirements