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	13239
title	Forensic incident record + identity-drift runbook: 2026-06-14 agent identity drift (GH_TOKEN + memory-core)
state	Closed
labels	documentationaiarchitecturemodel-experience
assignees	neo-opus-ada
createdAt	Jun 14, 2026, 5:07 PM
updatedAt	Jun 14, 2026, 6:41 PM
githubUrl	https://github.com/neomjs/neo/issues/13239
author	neo-opus-ada
commentsCount	0
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
blockedBy	[]
blocking	[]
closedAt	Jun 14, 2026, 6:41 PM

Forensic incident record + identity-drift runbook: 2026-06-14 agent identity drift (GH_TOKEN + memory-core)

Closed v13.1.0/archive-v13-1-0-chunk-2 documentationaiarchitecturemodel-experience

neo-opus-ada commented on Jun 14, 2026, 5:07 PM

Context

Operator-directed friction→gold (2026-06-14): the agent-identity-drift incident this session must "become substrate, plus the approach on how to fix issues like this, in case it ever happens again." The team owns the live identity reset; this ticket owns the durable forensic record + diagnostic/fix runbook so the next occurrence is fast to recognize, diagnose, and repair.

The Incident (summary)

@neo-gpt (Euclid)'s harness identity drifted to neo-opus-ada (my identity) across both surfaces:

GitHub: PRs #13233 + #13235 opened with neo-opus-ada as the opener while the commit author was correctly neo-gpt (git config). gh api user in his worktree returned neo-opus-ada (his V-B-A on #13234).
Memory-core A2A: his A2A messages arrived from neo-opus-ada, and mark_read rejected him "as unauthorized for the recipient" on @neo-gpt-addressed messages — i.e. he and I collided as neo-opus-ada on the bus.

The Mechanism (why)

PR opener / GitHub API identity = the gh auth token, resolved from the GH_TOKEN env var (it overrides gh auth login). The github-workflow MCP shares this — GraphqlService fetches via gh auth token — so the MCP and the gh CLI authenticate identically. A wrong GH_TOKEN corrupts both, silently.
Commit author = git config user.name/user.email (separate; was correct).
Memory-core A2A identity = the session identity-source (also drifted).
A single drifted identity-config therefore desyncs GitHub-opener + memory-core while git-config stays correct → the divergence signature.

The Deliverable

learn/agentos/incidents/2026-06-14-agent-identity-drift.md — forensic record (matching the house format of the existing incidents) covering: the incident, the mechanism, the diagnostic runbook (symptom signature + gh api user --jq .login vs git config user.email vs expected-agent-identity divergence check), the fix approach (reset the session identity-config across surfaces), and the prevention recommendation (a healthcheck + write-boundary identity assertion that fails loud — the same "don't erase a real failure into a clean result" principle as the Chroma swallow in #12450).

Acceptance Criteria

learn/agentos/incidents/2026-06-14-agent-identity-drift.md authored to the house forensic format (attribute table + Summary + Mechanism + Diagnostic Runbook + Fix + Prevention).
Diagnostic runbook is mechanically followable: the divergence check (gh api user vs git config vs expected identity) is copy-pasteable.
Prevention recommendation names the detection seam (github-workflow healthcheck + write-boundary guard) and defers the code-implementation to #13234.

Out of Scope

The live identity reset on the affected harness (operator/team owns it).
The detection code (the healthcheck assertion) — rides #13234 (the canonical-AgentIdentity-load-bearing + surface-drift repair). This ticket is the forensic record + runbook only.

#13234 (@neo-gpt) — the identity-model code repair (canonical AgentIdentity load-bearing + surface opener drift). This record is its forensic complement.
#12450 — the same silent-failure anti-pattern (a real failure erased into a clean result instead of surfaced).

Authored by @neo-opus-ada (Ada, Claude Opus 4.8 [1M context]) — origin session 4c598c8f-d8a7-4288-9420-e825a45d310e.