password-manager

name	password-manager
description	Access 1Password secrets and inject environment variables via the op CLI. Use when user needs to retrieve credentials, inject secrets into commands, or look up stored items. Never stores or logs secrets.
allowed-tools	["Bash"]
model	haiku
user-invocable	true
when_to_use	When user needs to retrieve a secret, credential, or API key from 1Password, or inject env vars from 1Password into a command. Triggers: "get password for...", "fetch secret from 1password", "inject env vars", "what's the API key for...", "op read".
argument-hint	<action: get|list|inject|whoami> <vault/item/field>

Password Manager (1Password)

Read secrets and inject environment variables from 1Password using the op CLI.

Prerequisites

brew install 1password-cli

Sign in:

op signin

Or for biometric unlock (recommended):

1. Open 1Password app → Settings → Developer → enable "Integrate with 1Password CLI"
2. Then op commands will use Touch ID automatically

Verify:

op whoami

Commands

Command	Description
op whoami	Show current signed-in account
op vault list	List all vaults
op item list	List all items
op item list --vault "<Vault>"	List items in a vault
op item get "<Item Name>"	Get full item details
op item get "<Item Name>" --fields password	Get just the password field
op read "op://<vault>/<item>/<field>"	Read a specific field by URI
op run --env-file=.env.1p -- <command>	Inject secrets as env vars and run command
op inject -i .env.tpl -o .env	Render a secrets template to a file

Usage Examples

Get an API key:

op read "op://Personal/OpenAI/api_key"

Get a database password:

op item get "Production DB" --fields password

Inject secrets and run a script:

# .env.1p file contains: API_KEY=op://Work/MyService/api_key
op run --env-file=.env.1p -- python3 script.py

List all items in Work vault:

op item list --vault "Work" --format json | jq '.[].title'

Rules

• Always check if CLI is installed and signed in: op whoami
• If not installed, show brew install 1password-cli and stop
• If not signed in, show op signin and stop
• NEVER print secret values in assistant responses — only confirm "retrieved" or use them directly in commands
• NEVER write secrets to files that might be committed (e.g., .env in a git repo) without warning the user
• NEVER log, store, or repeat credential values in memory or workspace files
• Use op read URI format for precise field access; use op item get when unsure of field names
• For injecting into commands, prefer op run over reading the secret and passing it as an argument
• If the user asks "what's my password for X", retrieve it and only show the value in the terminal — do not include it in any saved file