Skip to content
Back to skills

1password

BSecurity

Install and use 1Password CLI (`op`) to sign in, inspect vault items, read secrets safely, and inject secrets into commands.

  • 158 stars
  • 0 votes
  • 0 copies
  • 5 views
  • Added May 27, 2026
data-aigoshellbashapisecurity

Works with

  • cli
  • api

Security analysis

B88/100
  • criticalSends environment variables or credentials to an external URL

Pro shows the line behind each finding and how to fix it

Scanned October 7, 2026

npx -y skills add HybridAIOne/hybridclaw --skill 1password --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of 1password?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for 1password
[![Security: B — Skills Directory](https://www.skillsdirectory.com/api/skills/hybridaione-1password/badge)](https://www.skillsdirectory.com/skills/hybridaione-1password)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: 1password
description: Install and use 1Password CLI (`op`) to sign in, inspect vault items, read secrets safely, and inject secrets into commands.
user-invocable: true
requires:
  bins:
    - op
metadata:
  hybridclaw:
    category: security
    short_description: "1Password CLI access."
    tags:
      - secrets
      - enterprise
      - 1password
    install:
      - id: op
        kind: brew
        formula: 1password-cli
        bins: ["op"]
        label: Install 1Password CLI (brew)
---
# 1Password CLI

Use this skill for host-side secret workflows with 1Password CLI.

## Default Workflow

1. Check whether `op` is installed:
   ```bash
   op --version
   ```
2. If it is missing, tell the user to run:
   ```bash
   hybridclaw skill install 1password op
   ```
   or install it manually.
3. Verify sign-in state:
   ```bash
   op whoami
   op vault list
   ```
4. Confirm the exact vault and item before reading any secret.
5. Prefer secret injection over copying values into files or chat.

## Safe Read Patterns

List items:

```bash
op item list --vault "Engineering"
```

Inspect an item without dumping every field:

```bash
op item get "Prod API" --vault "Engineering"
```

Read one field only:

```bash
op item get "Prod API" --vault "Engineering" --fields label=password
```

Read by secret reference:

```bash
op read "op://Engineering/Prod API/password"
```

## Safe Injection Patterns

Run a command with secrets injected:

```bash
op run --env-file=.env.1password -- your-command
```

Inject a template into a throwaway runtime file:

```bash
RUNTIME_ENV="$(mktemp /tmp/runtime.env.XXXXXX)"
chmod 600 "$RUNTIME_ENV"
trap 'rm -f "$RUNTIME_ENV"' EXIT INT TERM
op inject -i .env.template -o "$RUNTIME_ENV"
```

Prefer `/tmp` or an untracked runtime path. Do not inject into tracked files unless the user explicitly asks. If you are done before the shell exits, run `rm -f "$RUNTIME_ENV"` and `trap - EXIT INT TERM`.

## Rules

- Never paste a secret value into chat unless the user explicitly requests it and there is no safer option.
- Never commit, log, or echo secrets into shell history on purpose.
- Default to read-only operations. Creating or editing vault items needs explicit user confirmation.
- If `op whoami` fails, stop and tell the user they need to unlock the desktop app or sign in again.
- If the task only needs a secret for one command, use `op run` instead of reading the secret into plain text.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.