OEI CLI (@oei/cli)

Open Execution Intelligence (OEI) is an execution-aware developer safety protocol that parses, analyzes, scores risk, and enforces safety gates before developer commands execute.

Installation

npm install -g @oei/cli

System Diagnostic Verification

Verify installation health, local system environment, and developer tools:

oei doctor

Usage & Commands

1. oei analyze "<command>"

Analyze a developer command against the OEI Knowledge & Risk Engine without executing it.

Safety Guarantee: oei analyze is strictly analysis-only and NEVER executes any requested process.

oei analyze "npm install express"
oei analyze "git push origin main --force"
oei analyze "solana program deploy target/deploy/app.so"

2. oei exec "<command>"

Evaluate a command through the OEI Execution Gate and conditionally execute it.

oei exec "node --version"
oei exec "npm install express"
oei exec "solana program deploy app.so" --dry-run

3. oei context

Inspect workspace runtime environment, operating system, and Git repository state safely.

oei context

4. oei config

Manage OEI system and AI provider configuration settings persisted at ~/.oei/config.json.

oei config list
oei config get AI_PROVIDER
oei config set AI_PROVIDER mock

5. oei knowledge

Inspect registered knowledge sources and query official advisories.

oei knowledge sources
oei knowledge search "Solana"

Safety & Execution Security Model

• Analysis Non-Execution: oei analyze never spawns child processes.
• Strict Gate Policy:
  • ALLOW / SUGGESTION: Executes automatically (with suggestion banner).
  • WARN: Requires developer confirmation [y/N] (default N in non-interactive environments).
  • BLOCK: Exits with code 2 and NEVER executes.
• No BLOCK Bypasses: Passing --yes / -y on a BLOCK decision will NEVER bypass policy.
• Shell Injection Protection: Complex shell operators (&&, ||, ;, >, >>, |, <, $(), `) are safely intercepted and rejected with exit code 5.
• Zero Secret Exposure: Passwords, tokens, API keys, and environment variables are protected via SafetyGuard and strictly excluded from logs, outputs, and AI reasoning boundaries.
• Offline AI Boundary: Operates 100% offline by default using AI_PROVIDER=mock. AI only provides explanation summaries and never alters deterministic risk scores or execution permissions.

Exit Codes

• 0: Success (executed process exited 0, or dry-run complete)
• 1: Executed process exited non-zero / missing argument / invalid config
• 2: Command blocked by OEI Security Policy (BLOCK)
• 3: User declined confirmation prompt (WARN -> n)
• 4: Internal gate or analysis error
• 5: Unsupported / unsafe shell expression