> Markdown twin of https://adna.network/reference/specification/6-naming-conventions/
> Index: https://adna.network/llms.txt · Full corpus: https://adna.network/llms-full.txt
> State is a build-time snapshot generated 2026-08-30 (UTC); nothing here is live.

# 6. Naming Conventions — aDNA Specification

> **Scan**: Underscores not hyphens, `type_descriptive_name.md` pattern, ALLCAPS governance list, directory naming, `type_` prefix convention.

*Decisions: C3*

## 6.1 File Naming

Content files MUST use **underscores** for word separation. Hyphens MUST NOT be used in aDNA content files.

**Pattern**: `type_descriptive_name.md`

Examples:
- `mission_adna_standard.md`
- `session_{username}_20260211_120000_gap_analysis.md`
- `customer_acme_corp.md`
- `context_research_ancient_dna.md`

**Exception**: Code-adjacent files in the project content layer (not inside the triad) MAY use hyphens to respect ecosystem conventions (npm, pip, etc.).

**Exception**: Tool-generated files (e.g., `how/tasks/` with plugin-generated names) MAY retain their generated naming format.

## 6.2 Governance File Naming

Governance files MUST use ALLCAPS names. The exhaustive list:

- `CLAUDE.md`
- `MANIFEST.md`
- `STATE.md`
- `AGENTS.md`
- `README.md`

No other files SHOULD use ALLCAPS naming. This list MUST NOT be extended without a standard revision.

## 6.3 Directory Naming

Directories MUST use lowercase with underscores: `context_library/`, `missions/`.

**Exception**: `.agentic/` uses a dot prefix (embedded triad convention).

## 6.4 Type Prefix Convention

The `type_` prefix pattern is RECOMMENDED for aDNA content files. It enables sorting, filtering, and at-a-glance identification. Common prefixes:

| Prefix | Content |
|--------|---------|
| `mission_` | Missions (legacy: `plan_`) |
| `session_` | Session files |
| `template_` | Templates |
| `customer_` | Customer records |
| `context_` | Context library files |
| `idea_` | Backlog ideas |
| `skill_` | Skill procedures |

## 6.5 Rename Protocol

When a vault, project, or persona is **renamed**, the rename MUST, at rename-time, sweep the vault's own **live-routing governance files** (`CLAUDE.md`, `STATE.md`, `AGENTS.md`) of self-references to the **old** name. A vault whose routing files still point at its prior identity is out-by-event (OBE) residue — masked when a back-compat shim keeps the stale references resolving, which is exactly why the sweep is mandatory rather than incidental.

**Scope discipline** (the load-bearing rule): the sweep targets the **live-routing self-reference subset ONLY** — a file's own governance/routing prose that names the vault. It MUST NOT rewrite legitimate **historical** cross-references: provenance prose, session history, ADR lineage, and changelog entries are retained verbatim (archive-don't-delete, §15). A naive whole-vault grep over-counts the defect by sweeping this history; the rename recipe carries a **keep/strip classifier** to separate the two.

Recipe: `how/skills/skill_project_rename.md`. Decision: ADR-042.

---