> ## Documentation Index
> Fetch the complete documentation index at: https://prism.ntecdev.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# ADR-14: E1 closed — PRISM:LOCAL-START/END blocks are NOT part of BIOS v2

> The PRISM:LOCAL-START / PRISM:LOCAL-END marker-block feature proposed in my earlier Plan #3 Wave E1 is CLOSED as withdrawn. CLAUDE.md and AGENTS.md replicas hav

<Info>**Status:** `accepted` · **ADR-14** · Filed 2026-04-17</Info>

## Decision

The PRISM:LOCAL-START / PRISM:LOCAL-END marker-block feature proposed in my earlier Plan #3 Wave E1 is CLOSED as withdrawn. CLAUDE.md and AGENTS.md replicas have NO user-modifiable zones. On prism\_sync\_bios the replica is overwritten from SOR in full. No preservation logic, no markers, no scratchpad region inside BIOS files. If a project needs project-local content that survives sync, it belongs in PRISM.md (the app-layer file) — PRISM.md is the composable, per-project home for project-specific content. If a need arises for org-wide overrides, that is ORG.md (Ring 2.5 per ADR #12), NOT a LOCAL block inside BIOS.

## Rationale

E1 was proposed mid-session before the four-ring chain had been clarified. Frank pushed back: "I thought that was a function of org.md not prism.md." I walked the feature back with a YAGNI recommendation. Plan #3 got superseded by Plan #2 later the same session, and E1 should have died with it — but the superseded TODOs still carried "open question" text, and Donna's cold-start via `prism_start` read them as still-live blockers on TODO #55 (SPEC-016 `prism_sync_bios`). This ADR is the explicit kill-shot. BIOS files are immutable-per-version: they are either the SOR verbatim or a drift-warned replica, nothing in between. Project-specific content lives in PRISM.md. Org-specific overrides live in ORG.md. The BIOS layer does not mix scopes. This also materially simplifies `prism_sync_bios`: no marker extraction, no reinsertion logic, no edge cases around malformed markers. A replica sync is a file write.

## Rationale

*(none)*

## Alternatives Considered

Keep LOCAL blocks as a minimal sticky-note feature — rejected; adds complexity to sync logic for a use case that hasn't materialized. Defer the decision — rejected; deferral is what left Donna blocked. The question is either live or dead, and it's dead. Move LOCAL-equivalent feature to PRISM.md — not needed; PRISM.md is already per-project by design, so project-local content has a home without any new mechanism.