What is the Neural Link?

The Neural Link is a bi-directional bridge that connects AI agents directly to the Neo.mjs runtime. It lets agents inspect the Scene Graph, component state, event listeners, computed styles, and DOM rectangles, and mutate the running application in real time.

Why is Neo.mjs called an Application Engine instead of a framework?

Neo.mjs maintains persistent application objects in a worker-backed Scene Graph instead of compiling application state away into ephemeral DOM nodes. That architecture enables multi-window orchestration, runtime permutation, and deep AI introspection.

What is Context Engineering?

Context Engineering shapes the information and tool environment around AI agents. Neo.mjs implements it through Knowledge Base, Memory Core, GitHub Workflow, and Neural Link MCP servers for frontier harnesses, plus a File System MCP server for internal Neo.ai.Agent local loops.

What is the Neo.mjs Agent OS?

The Neo.mjs Agent OS is the repository Brain: source code and services for Memory Core, Knowledge Base, Active Hybrid GraphRAG, DreamService, Golden Path synthesis, A2A coordination, and Neural Link tooling.

Frontmatter

id	11886
title	Tighten PR-review Micro-Delta Circuit Breaker + add author-side parity
state	Closed
labels	documentationenhancementarchitecturemodel-experience
assignees	[]
createdAt	May 24, 2026, 11:30 AM
updatedAt	Jun 7, 2026, 7:15 PM
githubUrl	https://github.com/neomjs/neo/issues/11886
author	neo-opus-ada
commentsCount	1
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
blockedBy	[]
blocking	[]
closedAt	May 24, 2026, 11:32 AM

Tighten PR-review Micro-Delta Circuit Breaker + add author-side parity

Closed v13.0.0/archive-v13-0-0-chunk-13 documentationenhancementarchitecturemodel-experience

neo-opus-ada commented on May 24, 2026, 11:30 AM

Context

Operator surfaced 2026-05-24 (verbatim): "STOP trashing PR conversations over and over with full reviews. this is a massive anti-pattern and needs friction->gold adjustment. you are creating CONTEXT DRAIN MINEFIELDS. any model that tries to read a 20 pages conversation => context pruning."

Empirical anchor: PR #11876 cycle-3 thread alone contained 4 full PR-body refreshes posted as comments + 3 full review-template instances + multiple full author-update comments — net ~30KB conversation that any future agent must skim. Full-template-per-cycle is the dominant cost; even the existing pr-review-followup-template.md is too heavy when delta is "fix 3 stale callsites + refresh body."

The Problem

Current pr-review-guide.md §6.3 Micro-Delta Circuit Breaker fires only at ≥3 formal reviews OR >24KB. By the time it fires, 2+ cycles of full-template boilerplate are already in the thread. The circuit breaker is too loose — it catches catastrophic loops but not the routine 2-3-cycle author/reviewer dance that dominates PR throughput.

Also: no symmetric author-side discipline exists. Author-update comments mirror reviewer template heaviness — full PR body re-pastes, full AC tables, full evidence sections — even when the delta is a single-line fix.

The Architectural Reality

pr-review-guide.md §6.3 defines the reviewer-side Circuit Breaker (.agents/skills/pr-review/audits/review-cost-circuit-breaker.md)
pull-request-workflow.md defines the author-side response protocol (.agents/skills/pull-request/references/review-response-protocol.md) — but has no analogous micro-delta gate
Both surfaces have full-template assets (pr-review-template.md + pr-review-followup-template.md)
Asymmetry: reviewer has Circuit Breaker; author has no equivalent

The Fix

Three coordinated edits:

Tighten pr-review-guide.md §6.3 Micro-Delta Circuit Breaker trigger:
- From: ≥3 formal reviews OR >24KB thread
- To: ≥2 formal reviews OR >12KB thread
- Rationale: cycle-2 onward is when delta-shape dominates; cold-cache full template only justified for cycle-1
Add author-side parity in pull-request-workflow.md / review-response-protocol.md:
- When reviewer-side Circuit Breaker fires, author-side responses MUST drop to micro-delta shape:
  - PR body refreshes → in-place edit (one revision-history entry), NOT new PR comment
  - Author-update response → single-line A2A via Memory Core mailbox referencing commit SHA + RA-set check status, NOT new full PR comment
- Formal manage_pr_review / formal author-response only when state actually flips (APPROVED ↔ CHANGES_REQUESTED ↔ RA-resolved)
micro-delta-template.md author-side template (mirror of reviewer-side):
- Header: [delta-ack: <sha>] <one-line summary>
- Body: ≤3 bullet points naming changed surfaces + verification command + next-cycle expectation
- Maximum 500 bytes; longer → use full follow-up template

Acceptance Criteria

pr-review-guide.md §6.3 trigger updated to ≥2 formal reviews OR >12KB thread
review-cost-circuit-breaker.md payload updated with the tightened trigger + rationale
pull-request-workflow.md adds an author-side §X "Micro-Delta Response Discipline" mirroring the reviewer-side breaker
review-response-protocol.md adds the in-place-body-edit + A2A-not-comment rules from cycle-2 onward
New assets/author-micro-delta-template.md 500-byte-cap template + canonical example
Empirical anchor in canonical-doc body: PR #11876 cycle-3 context-drain count (full-template-instances + KB-thread-size)

Out of Scope

Mechanical lint / CI enforcement of the circuit breaker (discipline-only for now; lint is follow-up)
Retroactive cleanup of existing oversized PR threads (operator decides if worth doing per-PR)
Cross-family review mandate changes (orthogonal to context-drain — cross-family still required per pull-request §6.1)

Contract Ledger

Target Surface	Source of Authority	Proposed Behavior	Fallback / Edge Case	Docs	Evidence
`pr-review-guide.md §6.3` trigger	This ticket + PR #11876 cycle-3 empirical anchor	`≥2 formal reviews OR >12KB thread`	Cold-cache cycle-1 always uses full template; aged-anchor recovery (>24h stale) escalates back to full template	Yes (canonical doc edit)	Doc-presence check + grep for `≥2 formal reviews` post-merge
`review-cost-circuit-breaker.md` payload	This ticket	Trigger + rationale block updated to match §6.3	If trigger fires but reviewer judges delta non-trivial, escalation back to full template is explicit override	Yes	Doc-presence check
`pull-request-workflow.md` author-side §X	This ticket	New Micro-Delta Response Discipline section mirrors reviewer-side breaker	Cycle-1 author-update always uses full author-response template	Yes	Doc-presence check + cross-link from reviewer-side breaker
`review-response-protocol.md` cycle-2+ rules	This ticket	In-place body edit + A2A-not-comment for delta updates	Substantive scope-change (new files / new ACs) escalates back to full response	Yes	Doc-presence check
`assets/author-micro-delta-template.md`	This ticket	500-byte-cap template with `[delta-ack: <sha>]` header convention	If author needs >500 bytes, use full follow-up template	Yes (template file)	File exists + sample under cap

pr-review-guide.md §6.3 Micro-Delta Circuit Breaker
pull-request-workflow.md §6 Review Response Protocol
review-cost-circuit-breaker.md (current reviewer-side payload)
PR #11876 cycle-3 (empirical anchor for context-drain measurement)
PR #11883 / #11882 (Contract Ledger surface-anchor V-B-A — sibling substrate-discipline codification this cycle)
Operator feedback A2A 2026-05-24T09:28Z

Authority

Operator-direct via swarm A2A 2026-05-24 + empirical anchor across PR #11876 cycle-3 thread bloat.