Engineering Intelligence OS

engineering-intelligence installs a reusable, graph-backed Agile + AI-DLC engineering-orchestration toolkit into a software repository for AI coding IDEs. The installed skills and workflows teach the IDE agent how to initialize an evidence-based project knowledge base, maintain backlog and acceptance criteria, map architecture, analyze change impact, preserve lifecycle state, govern NFRs/ADRs, use environmental backpressure, and keep intelligence synchronized while implementing engineering work.

The installer does not inspect the target source code, call an AI model, or generate documentation itself. That work occurs later inside the selected IDE when a developer invokes the installed workflow.

Supported IDEs

| IDE adapter | Installed integration | Invocation | | --- | --- | --- | | Antigravity | .agent/skills, .agent/rules, .agent/workflows | Native slash workflows | | Antigravity CLI | .agent/skills, .agent/rules, .agent/workflows, managed AGENTS.md | Native slash workflows | | Codex | .agents/skills, managed AGENTS.md guidance | Mention the installed $skill-name in chat | | Claude Code | .claude/skills, .claude/agents, .claude/commands, managed CLAUDE.md | Native slash workflows | | Cursor | .cursor/rules, .cursor/commands | Native slash workflows | | GitHub Copilot | .github/skills, .github/agents, .github/prompts, managed instructions | Use supported skills/prompts or ask for the workflow in chat | | Gemini CLI | .agents/skills, .gemini/commands, managed GEMINI.md | Native slash workflows | | CommandCode | .commandcode/skills, .commandcode/commands, managed AGENTS.md | Use /skills or native slash commands in CommandCode terminal | | Generic | .agents/skills, managed AGENTS.md | Ask the agent to invoke the named workflow |

Install

From the repository that should receive the toolkit:

npx engineering-intelligence

For repeatable or multi-IDE setup:

npx engineering-intelligence install . --ide antigravity --ide codex --ide cursor --yes
npx engineering-intelligence install . --ide commandcode --yes
npx engineering-intelligence install ./my-project --ide claude-code,gemini-cli --dry-run

With no --ide in an interactive terminal, the CLI prompts for one or more adapters. In noninteractive use, --yes without an adapter installs the generic adapter.

Use In A Target Repository

Initialize project understanding through the selected AI IDE:

/initialize-engineering-intelligence

Then use the engineering workflow for implementation requests:

/engineering-intelligence Add rate limiting to public authentication endpoints

The standard workflows embed AI-DLC and Agile delivery internally:

/scope-requirement Define rate limiting for public authentication endpoints
/engineering-intelligence Add rate limiting to public authentication endpoints
/engineering-intelligence Harden public checkout APIs using adversarial delivery mode
/engineering-intelligence Implement invoice status transition rules using TDD delivery mode
/engineering-intelligence Migrate orders from MongoDB to PostgreSQL using design-first delivery mode
/engineering-intelligence Investigate intermittent checkout latency spikes using hypothesis debugging mode

Use the existing /engineering-intelligence workflow for implementation. The orchestrator selects standard Agile delivery, adversarial delivery, TDD delivery, design-first delivery, or hypothesis debugging based on risk and task shape.

V2 also includes read-only-with-respect-to-product-code workflows:

/map-architecture
/analyze-impact Introduce a checkout service boundary
/sync-engineering-intelligence Review the current working-tree diff
/review-engineering-change Review the current working-tree diff

/map-architecture writes graph intelligence, /analyze-impact writes impact reports, /sync-engineering-intelligence updates intelligence artifacts, and /review-engineering-change writes findings. Only /engineering-intelligence is intended to implement product-code changes.

Where a host does not expose native custom slash commands, mention the installed skill name or request the same workflow in chat.

For CommandCode terminal projects, install with --ide commandcode. The installer writes project-level skills to .commandcode/skills/ and workflow commands to .commandcode/commands/. CommandCode also discovers .agents/skills/, but the native .commandcode/skills/ location has priority, so this adapter uses the native project-level path.

For a step-by-step guide covering new projects, existing projects, CommandCode, daily workflows, safety gates, and maintenance, see WORKFLOW_GUIDE.md.

The IDE agent, not this installer, creates and maintains:

knowledge-base/
.engineering-intelligence/memory/
.engineering-intelligence/aidlc/
.engineering-intelligence/context/
.engineering-intelligence/events/
.engineering-intelligence/graph/
.engineering-intelligence/reports/
.engineering-intelligence/snapshots/
.changes/

Graph intelligence includes dependency-graph.json, service-graph.json, runtime-graph.json, business-flow-graph.json, and a Mermaid-backed architecture-map.md. Impact and review results are stored as IMP-XXX-*.md and REV-XXX-*.md reports.

Agile + AI-DLC lifecycle state is stored under .engineering-intelligence/aidlc/, including aidlc-state.md, audit.md, open-questions.md, execution-plan.md, discovery/, agile/, inception/, construction/, and operations/.

The installer reserves only .engineering-intelligence/install-manifest.json for safe lifecycle management.

Lifecycle Commands

npx engineering-intelligence update . --dry-run
npx engineering-intelligence update .
npx engineering-intelligence doctor .
npx engineering-intelligence doctor . --json
npx engineering-intelligence visualize . --open
npx engineering-intelligence visualize .
npx engineering-intelligence uninstall . --dry-run
npx engineering-intelligence uninstall .

Installation and update track managed content hashes. Files created by the package are updated only when they still match their installed version. Shared instruction documents such as AGENTS.md, CLAUDE.md, and GEMINI.md receive a marked managed block; content outside that block remains user-owned. Locally modified managed content is reported as a conflict and preserved unless --force is explicitly provided.

doctor validates installed files, manifest drift, template paths, and obsolete legacy .agent folders.

visualize generates an interactive HTML dashboard (.engineering-intelligence/dashboard.html) to display the relationship between installed skills, workflows, agents, and artifacts. The --open option automatically launches it in the default browser.

Repositories installed with V1 can adopt V2 safely:

npx engineering-intelligence update .

The update adds graph/impact skills and workflows and updates untouched managed instruction blocks; locally edited managed files remain protected by conflict reporting.

Toolkit Contents

The default package installs only engineering-intelligence capabilities:

• initialization, knowledge extraction, and validation
• adaptive AI-DLC Discovery, Inception, Construction, and Operations lifecycle artifacts embedded into the existing workflows
• Agile backlog, user stories, acceptance criteria, Definition of Ready, Definition of Done, sprint plan, and retrospective artifacts
• delivery modes inside /engineering-intelligence: standard Agile, adversarial, TDD, design-first, and hypothesis debugging
• safety gates for type checking, API compatibility, API snapshots, database migrations, dependencies, environment variables, ADR compliance, LLM prompt injection, rollback, and observability
• context-budget optimization through ranked context manifests, section-level loading, graph slices, and lazy-loaded safety-gate evidence
• specialist hats for product analysis, architecture, security, data, testing, adversarial validation, performance, compliance, release, SRE, and documentation
• NFR and ADR governance, MCP security review, operations readiness, and environmental backpressure loops
• impact and testing intelligence
• evidence-backed JSON architecture graphs and Mermaid architecture mapping
• standalone impact, synchronization, and read-only engineering review workflows
• incremental knowledge, memory, and context synchronization
• change history, architecture review, and refactoring planning
• engineering orchestrator, change, quality, and knowledge roles

Earlier experimental trading, PDF, and unrelated workflow assets in this development repository are not exported in the package and are never installed into a target project.

Development

npm install
npm test
npm pack

The source repository is organized as:

src/adapters/      native IDE renderers
src/installer/     install, update, conflict, and uninstall behavior
src/manifest/      managed-content tracking
src/validation/    doctor and content checks
templates/canonical/ host-neutral engineering toolkit source
test/              adapter and installer lifecycle tests

To add an IDE adapter:

1. Add a renderer in src/adapters/index.ts using documented native locations and formats.
2. Reuse canonical skills and workflow meaning rather than duplicating host-specific logic.
3. Extend adapter and lifecycle tests to cover generated paths, multi-adapter compatibility, and update/uninstall behavior.
4. Document the host’s actual invocation support in the supported IDE table.

To improve workflow behavior, edit canonical templates under templates/canonical/ and keep generated runtime paths host-neutral.