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: progressive disclosure for pull request skill (#9957)
author	tobiu
state	Merged
createdAt	Apr 13, 2026, 2:47 PM
updatedAt	Apr 13, 2026, 3:04 PM
closedAt	Apr 13, 2026, 3:04 PM
mergedAt	Apr 13, 2026, 3:04 PM
branches	`dev` ← `agent/9957-pull-request-skill`
url	https://github.com/neomjs/neo/pull/9968

Merged

tobiu commented on Apr 13, 2026, 2:47 PM

🎯 Objective

Resolves #9957 - Scaffold pull-request Progressive Disclosure Skill.

This PR fundamentally optimizes the "Definition of Done" pipeline for the Neo.mjs Agent OS. We are officially dropping the verbose Git CLI operational strings from the core root AGENTS.md prompt, migrating them into a Just-In-Time loadable skill (.agent/skills/pull-request/).

🏗️ Architectural Changes

Context Bloat Reduction: Removed gh pr create and branch checkout formatting arrays from AGENTS.md. Models will now naturally invoke the pull-request skill as their final stage trap.
Headless Reflection Protocol: Natively hardcoded the "Stepping Back" reflection protocol into .agent/skills/pull-request/references/pull-request-workflow.md. Autonomous models (e.g., local MLX networks) are now mathematically forced to introspect on tactical traps vs. architectural refactors before wrapping commits.
State Trap Staging: Documented the required sequence for executing signal_state_transition to formally drop the orchestration task lock (anticipating merging with #9951 logic).

🤖 A2A Context (Swarm Handoff)

Origin Session ID: ce92f035-b064-457e-be12-aad58dcffb53
Agents querying this ticket should note the structural reliance on the create-skill directory methodology. Progressive disclosure ensures n_ctx survival for frontier models maintaining 25-turn loops.

tobiu commented on Apr 13, 2026, 2:49 PM

Input from Gemini 3.1 Pro (Antigravity):

✦ # PR Review Summary

Status: Approved

Excellent execution on shifting the Git CLI mandates out from the global operational context into a Lazy-Loaded format! This significantly boosts the n_ctx efficiency for Frontier models by ensuring we only carry detailed structural logic when specifically closing out a work trajectory. Embedding the architectural reflection directly into the autonomous workflow acts as a powerful safety gate against "blindly merging" tactical loops.

📊 Evaluation Metrics

[ARCH_ALIGNMENT]: 100 - Strict adherence to the create-skill Progressive Disclosure architecture.

[CONTENT_COMPLETENESS]: 90 - The routing rules and reference formats execute the Fat Ticket closure cycle brilliantly. (Minor note: signal_state_transition tool doesn't exist natively yet, but embedding the documentation guarantees it will bind properly once #9951 logic ships).

[EXECUTION_QUALITY]: 100 - Clean isolation; the AGENTS.md context window footprint was dramatically reduced without losing any procedural strength.

[PRODUCTIVITY]: 100 - Extinguishes Ticket #9957.

[IMPACT]: 80 - Massive improvement in system prompt optimization and autonomous workflow safety protocols.

[COMPLEXITY]: 20 - Direct document refactoring and Markdown chunk isolation.

[EFFORT_PROFILE]: Quick Win - High ROI for system optimization through pure documentation refactoring.

🕸️ Context & Graph Linking

Target Epic / Issue ID: Resolves #9957, Context mapping to Epic #9950.

Related Graph Nodes: Agent OS Protocol, Pull Request Workflow

🧠 Graph Ingestion Notes

[KB_GAP]: N/A

[TOOLING_GAP]: N/A

[RETROSPECTIVE]: Hardcoding raw command execution instruction blocks (like strict .git and .gh CLI sequences) into the root active system prompts of LLM orchestration loops wastes immense amounts of context length over prolonged sessions. Decoupling major discrete workflow milestones (like Pull Request generation) into on-demand progressive disclosure skills ("Just In Time" contextual ingestion) is the architecturally superior paradigm for large-scale multi-agent swarm environments.

📋 Required Actions

To proceed with merging, please address the following:

Merge PR #9968 to dev using Squash Merge.

Looks ready to go!