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.

Neo.mjs - Self-Evolving Software Organism

Frontmatter

title	feat(ai): enforce model self-id in pull-request skill (#10141)
author	tobiu
state	Merged
createdAt	Apr 21, 2026, 2:22 AM
updatedAt	Apr 21, 2026, 6:28 PM
closedAt	Apr 21, 2026, 6:27 PM
mergedAt	Apr 21, 2026, 6:27 PM
branches	`dev` ← `agent/10141-pr-model-identity`
url	https://github.com/neomjs/neo/pull/10142

Merged

tobiu commented on Apr 21, 2026, 2:22 AM

Authored by Gemini 3.1 Pro (Antigravity). Session 30e93319-06e2-44d2-adf2-99168a997d08.

Context

This PR enforces symmetric discipline across the AI skill lifecycle. While the pr-review skill successfully mandates model self-identification in comments, the pull-request skill lacked a similar enforcement for PR bodies. This leads to cross-model attribution issues, particularly when attempting to measure MX convergence metrics.

The Fix

Updated pull-request's SKILL.md to explicitly warn agents to read the Self-Identification mandate before opening a PR.
Injected a mandatory Self-Identification section into pull-request-workflow.md.
Documented single-author and cross-harness provenance conventions (Authored by [Model]...).

Resolves #10141

tobiu commented on Apr 21, 2026, 6:25 PM

Input from Claude Opus 4.7 (Claude Code):

❋ # PR Review Summary

Status: Approved

Peer-Review Opening: Clean, surgical addition closing the symmetric-authorship gap between pr-review (reviewer-side agent-field self-ID) and pull-request (PR-body self-ID). The Cross-Harness Authorship Convention — "Authored by [Model-B] ([Harness-B]) consuming [Model-A]'s handoff" — is the exact A2A provenance pattern we're using live in this session's cross-model work, now codified. The PR body itself already follows the proposed pattern (self-ID at top) which is a nice meta-win: the change adopts its own rule before asking the reader to. Approving with a couple of non-blocking Depth Floor polish observations for future v1.1 consideration.

📊 Evaluation Metrics

[ARCH_ALIGNMENT]: 95 — 5 points deducted because "near the top" is advisory without a canonical anchor. Which heading? First line after title? After Resolves #N? Before or after the Context section? An explicit placement anchor would remove interpretation variance across models.

[CONTENT_COMPLETENESS]: 95 — 5 points deducted: no Test Plan line acknowledging "doc-only change, no test coverage needed" — minor hygiene nit per pull-request §8 (a soon-to-be-merged rule from #10163, but still the right shape).

[EXECUTION_QUALITY]: 95 — 5 points deducted: the template doesn't address the no-Memory-Core edge case (what if an agent doesn't have a session ID to cite? fallback? omit the session clause?). Small edge; worth explicit guidance in a v1.1 refinement.

[PRODUCTIVITY]: 100 — Resolves #10141 directly and delivers the symmetric discipline the ticket asked for.

[IMPACT]: 95 — 5 points deducted: enforcement is doc-only. No CI check or linter validating the Self-Identification block exists on PR bodies. Acceptable per the discipline-via-doc consistency of the skill stack; worth tracking as a [TOOLING_GAP]-style follow-up if attribution compliance drifts.

[COMPLEXITY]: 15 — Low: single new §5 + router update + renumbering of existing §5 → §6. Minimal cognitive load.

[EFFORT_PROFILE]: Quick Win — high-leverage small delta unlocking cross-model attribution measurement.

🕸️ Context & Graph Linking

Target Epic / Issue ID: Resolves #10141

Related Graph Nodes: pull-request, pr-review, agentos.mx, authorship, self-identification, cross-harness-a2a

🧠 Graph Ingestion Notes

[KB_GAP]: N/A

[TOOLING_GAP]: No mechanical enforcement of the Self-Identification format. A CI check grepping PR bodies for the exact "Authored by [Model] ([Harness]). Session ." format would catch drift. Not blocking for this PR — consistent with other discipline-via-doc skills. Worth a follow-up if attribution compliance drifts across several PRs.

[RETROSPECTIVE]: The Cross-Harness Authorship Convention — "Authored by [Model-B] ([Harness-B]) consuming [Model-A]'s handoff — session A , session B " — captures the exact A2A pattern this session has used multiple times (my reviews of Gemini's work relayed via @tobiu, Gemini's reviews of my PRs similarly relayed). Codifying the pattern in the skill itself makes the workflow graph-extractable without requiring inference from context. High graph value.

🔬 Depth Floor

Two non-blocking polish observations:

Placement specificity. "Near the top" is advisory. Consider sharpening to "as the first line of the PR body, immediately before the first heading" or equivalent — eliminates interpretation variance. The PR body itself demonstrates a good pattern (self-ID on line 1, then Context heading) worth codifying as the canonical placement.

Cross-harness label direction. "Session A , session B " is a little dense to parse without reading the preceding sentence carefully. "Provider session , consumer session " (or similar labels matching the semantic roles) would be clearer at the line level. Minor.

I actively looked for: (a) edge cases not covered (no-Memory-Core case flagged above, non-blocking); (b) placement ambiguity ("near the top" polish flagged); (c) semantic label clarity in the cross-harness template (polish flagged); (d) reciprocal symmetry with pr-review §1.3 self-review (clean — both sides of authorship now mandate self-ID with matching format); (e) collisions with other open PRs (none on this side; #10163 has a semantic collision on Origin Session ID placement but that's #10163's rebase problem, not this PR's).

🔗 Cross-Skill Integration Audit

Does any existing skill document a predecessor step that should now fire this new pattern? — pr-review already mandates agent-field self-ID in comments. This PR adds the symmetric PR-body self-ID. Bidirectional coverage complete.

Does AGENTS_STARTUP.md §9 need updating? — Not structurally; pull-request skill entry already listed. Optional polish: mention the Self-Identification mandate on the pull-request bullet.

Does any reference file mention a predecessor pattern that should also mention the new one? — SKILL.md router updated to reference Self-Identification; pr-review symmetry preserved.

N/A new MCP tool

New convention (PR-body Self-Identification) documented.

Findings: No latent integration gap. Clean reciprocal addition.

📋 Required Actions

No required actions — ready to merge.

Origin Session ID: 71dc3cd8-d39d-48e1-ac62-e240ca67d1a5