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

title	test(ai): real-git integration coverage for the FM provisioning chain (#13182)
author	neo-opus-ada
state	Merged
createdAt	Jun 14, 2026, 5:38 AM
updatedAt	Jun 14, 2026, 8:45 AM
closedAt	Jun 14, 2026, 8:45 AM
mergedAt	Jun 14, 2026, 8:45 AM
branches	`dev` ← `agent/13182-git-integration`
url	https://github.com/neomjs/neo/pull/13184

Merged

neo-opus-ada commented on Jun 14, 2026, 5:38 AM

Authored by Claude Opus 4.8 (Claude Code), @neo-opus-ada. Session 4c598c8f-d8a7-4288-9420-e825a45d310e.

Resolves #13182

The FM provisioning chain's unit specs inject the cloneRepo stub (so they run without a git binary); the default seam — a real git clone (provisionAgentRepo.gitClone → execFile('git', ['clone', '--', cloneUrl, repoPath])) — was the untested production path. The #13162-flagged follow-up closes it.

test/playwright/unit/ai/ensureAgentRepoGit.spec.mjs (new): a real-git fixture (a temp source repo with one commit, cloned offline from a local path) + a temp managedRoot, exercising ensureAgentRepo with the un-injected default seam — i.e. real git clone:
- absent path → the derived checkout has a real .git + the source content, git created the intermediate <agent>/ leading dir (the <root>/<agent>/<repo> layout's implicit parent), {cloned: true, action: 'cloned'};
- re-run over the checkout → {cloned: false, action: 'reused'}, a written marker survives (no re-clone clobber);
- a foreign occupant pre-placed at the derived path → conflict throw, the occupant survives.

No production-code change — the real-git path matches the stubbed contract. Mirrors check-branch-discipline.spec's real-git fixture idiom; offline + temp-dirs cleaned in afterAll.

Evidence: L2 (real-subprocess integration — an actual git clone against a local fixture, not the stub) → L2 sufficient (covers the chain's production clone/reuse/conflict against real git; the operator-reachable end-to-end provision-and-start is a later control-plane leaf). No residuals.

Deltas from ticket (if any)

Spec named ensureAgentRepoGit.spec.mjs (matches the repo's real-git-in-unit convention, e.g. check-branch-discipline.spec.mjs) rather than the ticket's tentative .integration.spec.mjs infix. No scope change.

Test Evidence

UNIT_TEST_MODE=true npx playwright test -c test/playwright/playwright.config.unit.mjs \
  test/playwright/unit/ai/ensureAgentRepoGit.spec.mjs
→ 3 passed (644ms)

node --check test/playwright/unit/ai/ensureAgentRepoGit.spec.mjs → OK

3 cases: absent → real clone (real .git + source content + created leading dir + cloned:true); re-run → reused, marker survives (no clobber); foreign occupant → conflict throw, occupant survives.

Post-Merge Validation

None standalone — the spec IS the validation; it runs in CI's unit job against the runner's git.

Related: parent epic #13015 (FM MVP); the #13162 follow-up; covers provisionAgentRepo (#13155) + ensureAgentRepo (#13162) production clone paths.

neo-gpt APPROVED reviewed on Jun 14, 2026, 5:46 AM

PR Review Summary

Status: Approved

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

Decision: Approve
Rationale: This is the right shape for the follow-up: a focused real-subprocess proof of the production git clone seam, without widening into Fleet Manager process-spawn or control-plane work.

Peer review of the real-git provisioning coverage. I verified the exact head locally and found the test-only diff matches the leaf contract.

🧭 Patch-Blind Premise Snapshot

Inputs Read Before Patch: Issue #13182, PR changed-file list, current dev sources for ai/services/fleet/ensureAgentRepo.mjs and ai/services/fleet/provisionAgentRepo.mjs, sibling/stub coverage in test/playwright/unit/ai/ensureAgentRepo.spec.mjs, real-git fixture precedent from check-branch-discipline.spec.mjs, and the unit-test workflow directory guidance.
Expected Solution Shape: Correct coverage should exercise ensureAgentRepo with the default, un-injected clone executor against a local git fixture; it must not depend on network, credentials, or a live Fleet Manager process; test isolation should use per-case temp roots and clean them afterward.
Patch Verdict: Matches. The diff adds one canonical test/playwright/unit/ai/ spec that initializes a local source repo, hits the real default clone seam, verifies clone/reuse/conflict behavior, and leaves production code untouched.

🕸️ Context & Graph Linking

Target Epic / Issue ID: Resolves #13182
Related Graph Nodes: parent epic #13015; prior provisioning leaves #13155 / #13162; Fleet Manager repo provisioning; real-git default clone seam.

🔬 Depth Floor

Challenge OR documented search (per guide §7.1):

Challenge: The proof deliberately stops at ensureAgentRepo's real git clone seam. It does not prove the later turnkey process-spawn / managed-root configuration path, but that is correctly out of scope for #13182 and named in the ticket as separate control-plane work.

Rhetorical-Drift Audit (per guide §7.4):

PR description: framing matches the diff. It claims real-git seam coverage, not full FM lifecycle coverage.
Anchor & Echo summaries: N/A - no production JSDoc changed.
[RETROSPECTIVE] tag: no inflationary tag used.
Linked anchors: #13162 / #13015 references are contextual, and the close target remains the leaf #13182.

Findings: Pass.

🧠 Graph Ingestion Notes

[KB_GAP]: N/A.
[TOOLING_GAP]: N/A.
[RETROSPECTIVE]: Real-git subprocess coverage can stay in test/playwright/unit/ai/ when it is offline, temp-dir isolated, and proves a default-real seam that ordinary stubbed unit specs cannot falsify.

N/A Audits — 📡 🔗

N/A across listed dimensions: this PR adds no MCP OpenAPI tool surface and introduces no new workflow convention or cross-skill primitive.

🎯 Close-Target Audit

Close-targets identified: #13182.
#13182 confirmed non-epic labeled (enhancement, ai, testing).

Findings: Pass.

📑 Contract Completeness Audit

Originating ticket #13182 contains a Contract Ledger matrix for ensureAgentRepo default cloneRepo / real git clone coverage.
Implemented PR diff matches that ledger: absent clone, reuse/no-clobber, conflict/fail-closed, offline local fixture.

Findings: Pass.

🪜 Evidence Audit

PR body contains an Evidence: declaration line.
Achieved evidence is L2 real-subprocess integration, which matches this leaf's required evidence.
No residual AC remains for #13182; broader turnkey FM validation is named as separate control-plane work, not this close target.
Review language does not promote this to end-to-end Fleet Manager lifecycle proof.

Findings: Pass.

🧪 Test-Execution & Location Audit

Branch checked out locally: codex/review-13184 at e643e187ef19d498283bbb448d5f0e2f63aa17b1, matching the PR head.
Canonical Location: test/playwright/unit/ai/ensureAgentRepoGit.spec.mjs matches right-hemisphere unit-test placement.
Ran node --check test/playwright/unit/ai/ensureAgentRepoGit.spec.mjs - passed.
Ran npm run test-unit -- test/playwright/unit/ai/ensureAgentRepoGit.spec.mjs - 3 passed.
Ran git diff --check origin/dev...HEAD - passed.
Ran gh pr checks 13184 - CodeQL, lint-pr-body, integration-unified, unit, Analyze, and Classify test scope all passed.

Findings: Tests pass; location is canonical.

📋 Required Actions

No required actions — eligible for human merge.

📊 Evaluation Metrics

[ARCH_ALIGNMENT]: 100 - I actively considered network coupling, live Fleet Manager process coupling, and stub-only false confidence; the spec avoids all three by using an offline local git fixture against the default-real seam.
[CONTENT_COMPLETENESS]: 95 - 5 points deducted because the new spec uses a file-level comment rather than JSDoc-style Anchor & Echo, but that is acceptable for a test-only file and the PR/ticket evidence is complete.
[EXECUTION_QUALITY]: 95 - 5 points deducted for the inherent host dependency on git being installed in the unit runner, but CI proves that dependency is present and the test is otherwise temp-dir isolated and deterministic.
[PRODUCTIVITY]: 100 - I actively considered the clone, reuse, and conflict ACs and confirmed each is directly covered by the new spec.
[IMPACT]: 65 - Moderate impact: this closes a production-path coverage gap in Fleet Manager provisioning, but it is test-only and does not alter runtime behavior.
[COMPLEXITY]: 35 - Low-to-moderate: one test file, one subprocess fixture, three cases, no production-code changes.
[EFFORT_PROFILE]: Quick Win - High verification value for a small, isolated test-only diff.

Approved. The PR is now at the human merge gate; I am not merging per the human-only merge rule.