← ClaudeAtlas

password-generatorlisted

Generate cryptographically secure passwords or diceware-style passphrases via a bundled CSPRNG script — never by having the model type out a "random-looking" string. Use when the user asks to generate/create a password, passphrase, secret, PIN, or random token, or to check the strength of one.
vslipchenko/ai · ★ 0 · Data & Documents · score 68
Install: claude install-skill vslipchenko/ai
# Password Generator Produce passwords and passphrases whose randomness actually comes from a cryptographically secure random number generator, not from the model. **Why a script and not just "think of a random string": an LLM's output is not a source of entropy** — sampled tokens are statistically biased and, given the conversation, effectively predictable. A "random-looking" string typed by the model is not a secure password no matter how random it looks. This skill always shells out to `scripts/generate_password.{py,mjs}`, which use a real CSPRNG (Python's `secrets` module / Node's `crypto.randomInt`), for every character or word chosen. `<SKILL_DIR>` below is this skill's own announced base directory (the path shown when this skill loads, ending in `.../skills/password-generator`). ## Step 1 — Resolve mode and parameters Two modes: | Mode | What it produces | Good for | |---|---|---| | `password` | random characters from chosen classes (lower/upper/digits/symbols) | sites/systems with fixed complexity rules | | `passphrase` | N words from the bundled 7,776-word EFF diceware list | memorable, typed-by-hand secrets, master passwords | Default to `password` unless the user says "passphrase," "memorable," "words," or similar, or the target system disallows symbols (in which case a passphrase is often the better fit than a stripped-down password). Ask only if genuinely ambiguous; otherwise use sane defaults and mention them in the result so the user can redirect: - **