Inline Secret Block

An Obsidian plugin that encrypts secrets inside your notes using AES-256-GCM and a passphrase — both fenced code blocks (multi-line) and single-line inline code spans. Useful when your vault is read by sync services, backup tools, or AI agents and you do not want passwords, tokens, or other secrets sitting in plaintext.

You write a secret block (or a `secret …` inline span), the plugin asks for a passphrase as soon as you move away from it, and rewrites it as a secret-lock form whose body is opaque ciphertext. In reading view and Live Preview the encrypted secret renders as something you can reveal or copy in place. Multiple passphrases can be in use at the same time — the plugin remembers each one you have entered in the current session and always lets you choose which key to use for a new secret.

How it works

You wrap the sensitive content in a fenced block with the language secret:

The login for service X:

```secret
login: alice
pass: hunter2
token: gh_xxxxxxxxxxxxxxxxxx
```

The moment the closing fence is in place, a modal opens. If you have not used any passphrase yet, it just asks for one. If you have already used some in this session, it lists them (labelled like a...2 9f1c0b — first character, last character, and the first 6 hex characters of the passphrase hash) so you can pick an existing key, or enter a new one. Pressing Submit replaces the block with:

The login for service X:

```secret-lock
QmFzZTY0ZW5jb2RlZHBheWxvYWQuLi4=
```

Pressing Cancel (or Esc) leaves the block as plaintext. If you change the body afterwards, the prompt comes back; if you do not touch it, the plugin will not nag you again about that exact block.

In reading view, each secret-lock block becomes a card with a lock icon and Show / Edit / Copy buttons. Press Show to decrypt and reveal the content; press Hide to clear it from the DOM again. Copy sends the plaintext to the clipboard without revealing it on screen. Edit decrypts the block back to a plain secret block right in the file so you can change it, and auto-encrypt re-locks it as soon as you click outside or stop typing — with the same key it was encrypted with, without asking for a passphrase again. The plugin tries every passphrase it has in memory automatically — you only see the password modal if none of them fit.

When you already have keys in memory, the chooser is shown every time you encrypt a new secret, so you stay in control of which key it uses. The last key you used in the current file is listed first and pre-focused, so reusing it is a single click (just press Enter), while picking a different key or typing a brand-new passphrase is always one step away. Re-locking an existing secret after Edit is the exception: it reuses that secret's own key automatically and never prompts.

Single-line (inline) secrets

For short secrets you do not need a whole fenced block. Wrap the value in an inline code span prefixed with secret :

password: `secret hunter2`

As soon as you move the cursor off that span, the plugin prompts for a key and rewrites it as:

password: `secret-lock QmFzZTY0Li4u`

In both Live Preview and reading view the secret-lock span renders as a compact chip with a lock icon and three buttons:

• Show (eye) — reveal the value inline; click again to hide it.
• Copy — copy the plaintext to the clipboard (a "Copied" notice appears; the value is never shown on screen). Clicking the chip body copies too.
• Edit (pencil) — decrypt the span back to a plain secret span so you can change it. Auto-encrypt re-locks it with the same key — no passphrase prompt (see below).

The Always show secret preview setting applies to inline chips as well: when enabled, a chip whose key is already in memory reveals automatically.

Inline secrets deliberately have no custom label/alias (unlike fenced blocks) to keep them simple. In Live Preview, move the cursor into the chip to edit the raw secret-lock text directly.

If you need to convert many secrets in one go (e.g. before bulk-editing a note), run Decrypt secret-lock blocks in current note to revert all secret-lock blocks and inline spans in the active note back to secret.

Commands

Wrong passphrase on the decrypt command is a no-op: the file is not changed and you get a notice. Same for corrupted blocks.

In a reading-view card, a wrong passphrase shows an inline error and a Try another passphrase button — use it when the block was encrypted with a key the plugin does not have in memory yet.

Settings

The plugin has a single setting under Settings → Community plugins → Inline Secret Block:

• Always show secret preview — when enabled, every secret-lock block and inline chip whose key is already in memory reveals automatically, without you having to press Show. Secrets whose key is not in memory stay hidden as usual. Off by default.

Cryptography

• AES-256-GCM for confidentiality and integrity. A wrong passphrase or any tampering with the ciphertext is detected by GCM's authentication tag.
• PBKDF2-SHA-256, 250 000 iterations to derive the key from your passphrase.
• Per-block random 16-byte salt and 12-byte IV. Both are prepended to the ciphertext and the whole salt ‖ iv ‖ ciphertext+tag blob is base64-encoded — that is what you see between the fences.
• All primitives come from the platform Web Crypto API (crypto.subtle). No third-party crypto dependencies.
• Derived keys are cached in memory keyed by (passphrase fingerprint, salt) so opening a note with many blocks does not re-run PBKDF2 each time.

Threat model

What the plugin does protect:

• The content of the secret in the file on disk. Whoever reads the .md file — sync provider, cloud backup, AI agent that crawls the vault, another Obsidian plugin — sees ciphertext, not your password.

What the plugin does not protect, and these are real:

• Plaintext residues after in-place encryption. Once you have typed the secret into a secret block, the plaintext may already exist in:

  • the editor's undo history (cleared when you reload the file),
  • Obsidian's File Recovery snapshots,
  • your sync provider's version history,
  • filesystem-level backups (Time Machine, restic, etc.).

  Recommendation: enter the secret in a secret block and run encrypt before the first sync of the file. If the file has already been synced with plaintext, assume the plaintext is out there forever and rotate the affected credential.

• Plaintext on screen. Once you press Show (or Copy), the secret is in the DOM and the clipboard. Anything that can read your DOM (other plugins, screen readers, accessibility tools, screenshots) can read it.

• No passphrase recovery. Forget the passphrase, lose the data. There is no escrow, no hint, no recovery email.

• The cryptography has not been independently audited. The primitives are standard and used as documented, but you should treat this as a best-effort tool, not a vetted security product. If your threat model requires audited crypto, use a dedicated password manager.

• The key picker leaks the first and last character of each passphrase. The label shown next to each in-memory key is <first>...<last> <hash6>, so an attacker who briefly sees the picker learns two characters and ~24 bits of the hash. This is an intentional UX trade-off so you can tell keys apart at a glance; it makes short or low-entropy passphrases noticeably easier to brute-force. Use long, high-entropy passphrases.

Passphrases are held in memory for the duration of your Obsidian session (or until you run Forget all passphrases). They are never written to disk and never sent over the network. The plugin does not write any persistent state at all.

Installation

Manual install

1. Download main.js, manifest.json, and styles.css from the latest GitHub release.
2. Copy them into <Vault>/.obsidian/plugins/inline-secret-block/ (create the folder).
3. In Obsidian, open Settings → Community plugins, reload the plugin list, and enable Inline Secret Block.

Build from source

git clone https://github.com/vnrtmnv/obsidian-inline-secret-block
cd obsidian-inline-secret-block
npm install
npm run build

main.js is produced at the repository root. Copy it together with manifest.json and styles.css as above.

One-shot local install script

For convenience there is a build_local.sh script at the repo root that builds the plugin and stages the install-ready files in one go:

./build_local.sh

What it does:

1. Runs npm install if node_modules/ is missing.
2. Runs npm run build (type check + esbuild production bundle).
3. Reads the plugin id from manifest.json and stages main.js, manifest.json, and styles.css into output/<plugin-id>/.

After the script finishes, copy output/inline-secret-block/ into <Vault>/.obsidian/plugins/ and reload Obsidian. The output/ directory is gitignored, so you can run the script as often as you like without polluting the working tree.

License

MIT. See LICENSE.