Stenos is an early-stage project. Security fixes are applied to the latest release on the default branch. Older versions are not maintained.

Please report security issues privately, not through public issues.

• Preferred: open a private security advisory with the "Report a vulnerability" button on the repository's Security tab.
• Alternatively, email the maintainer at stivenagostingjekaj@gmail.com.

Please include steps to reproduce, the affected version or commit, and the impact as you understand it. You can expect an initial response within a few days. Once a fix is ready it will be released, and your report will be acknowledged unless you prefer to remain anonymous.

Stenos records the audio of everyone in a voice channel and writes it to disk as text. Treat the following as known and documented properties rather than vulnerabilities.

• Recorded audio is buffered in memory unencrypted while a call is in progress, and transcripts are written to transcripts/ with no encryption at rest. Anyone who can read that directory can read the transcript.
• The bot token grants full control of the bot. It is read from the environment or a .env file, which is gitignored. Protecting it is the operator's responsibility.
• Anyone who can use the slash commands can start a recording. Stenos does not implement its own permission model; restrict the commands through Discord's own channel and role permissions if you need to.
• Recording is announced, never silent. The start and stop messages are deliberate and non-ephemeral. A request to add a silent mode is a feature request that will be declined, not a security report.

Model weights are downloaded from Hugging Face on first use. That is the only network destination beyond Discord itself, and no audio or transcript is ever sent anywhere.

• The lack of receive-side DAVE support in py-cord. That is an upstream limitation, documented in the README, not a vulnerability in Stenos.
• Denial of service caused by recording an extremely long call. Buffering is bounded by the host's memory, and the operator chooses when to stop.