llm-secrets is a security tool. We take vulnerability reports seriously.

Do not open a public issue for security vulnerabilities.

Email security@adjoint.uk with:

• A description of the issue
• Steps to reproduce (or a proof of concept)
• The affected version (llms --version)
• Your assessment of impact

You should receive an acknowledgement within 72 hours. We will work with you to confirm, fix, and disclose.

In scope:

• Anything that causes a secret to be exposed to the LLM context, stdout, logs, or an unauthorised process
• Bypasses of the policy engine (once shipped in v0.3)
• Bypasses of lease expiry or the killswitch (once shipped in v0.4)
• Cryptographic flaws in the on-disk format
• Memory hygiene issues (plaintext lingering after use)

Out of scope:

• Issues that require root on the user's machine (game over already)
• Side-channel attacks against the underlying age / ed25519-dalek crates (report upstream)
• Social engineering / phishing of users

Pre-1.0. We patch the latest released version only.