Security

A userbot acts as your account, so Uroboros guards it on three fronts: who can run commands, what code can be installed, and what that code can do at runtime.

This is not a sandbox. Third-party modules run in the bot's process, and a determined author can get around the checks. They stop typical malicious code and accidents, not arbitrary Python. Install modules you trust.

Who can run commands

Your own (outgoing) messages always. Commands in other people's messages run only if the sender has enough access:

Level	Who	Default commands
owner	this account and users added with .owner add	everything, including .e, .t, .dlm, .ulm, .restore, .update
sudo	added with .sudo add	most commands
support	added with .support add	help, .ping, .info
everyone	anyone	only commands explicitly set to this level

Each level includes the ones below it. Change a command's level with .security <command> <level>; reset with .security <command> default. .security alone shows groups, overridden permissions, frozen and trusted modules.

Code scanning on install

Before .dlm, .lm, .uplm and .restore, Uroboros reads the module source and looks for:

• session access: client.session.save(), StringSession.save, auth_key, api_hash, .session files, uroboros.db;
• account takeover requests: terminating sessions, deleting the account, changing 2FA or phone, QR login, logout, spending stars, transferring gifts;
• hidden code: exec/eval of base64, zlib, marshal;
• Uroboros system settings: access rights, flood protection, inline bot token.

A module with dangerous code isn't installed until you confirm explicitly: the "Install anyway" button or -f (.dlm -f, .uplm -f, .restore -f). This applies to modules from connected repos too.

Suspicious things (environment variables, shell commands, file deletion, config.json, exec of a string, channel joins) are listed separately in the reply but don't block install.

A module can declare what it needs: # meta permissions: network, files. Permissions are shown before install, and undeclared usage triggers a warning. See Writing modules.

Module source and version

• Modules download over https only; redirects to http are rejected.
• A module from an unconnected source needs confirmation showing its origin and line count.
• GitHub modules are pinned to a commit: the file is fetched by SHA, and the confirmation shows which commit.
• .uplm shows each module's diff and updates after confirmation.
• Uroboros stores sha256 of installed module files. If a file was changed outside Uroboros and now contains dangerous code, it won't load at startup.
• # requires: accepts only package names with versions: no URLs or pip options.

Runtime protection

A third-party module can't:

• read self.client.session;
• open, delete or rename the session, config.json or uroboros.db, or pass them to shell commands;
• send requests that hijack the account or spend money (the same ones the install scan looks for).

The attempt raises PermissionError, nothing happens, and Saved Messages gets a notice naming the module and the action.

Restrictions are lifted for a module whose dangerous code you confirmed on install, and for one you trust explicitly: .security trust <module>. Revert with .security untrust <module>.

Flood protection

Telegram requests from third-party modules are counted. A module that sends too many (default: more than 60 in 30 seconds) is frozen for 5 minutes and you get notified. Configure with .security flood <requests> <seconds> <freeze>, disable with .security flood off, unfreeze with .security unfreeze <module>.

Data

The session, config.json (api_id and api_hash) and the database live in data/. That's access to your account: never publish or share this folder. .backup saves the database and modules, but not the session or config.json.