# Security policy LearnHarness stores API keys, sign-in tokens, chats, and learning records on your machine, and its agent reads files and fetches web content on your behalf. If you find a way to abuse any of that, please tell us privately so a fix can ship before the details are public. ## Reporting a vulnerability **Do not open a public issue, pull request, or Discord post for a security problem.** Report it through GitHub's private vulnerability reporting: [Security → Report a vulnerability](https://github.com/amosblomqvist/learnharness/security/advisories/new). If that page is unavailable, message a maintainer on [Discord](https://discord.gg/5tCHHQEB2e) and ask for a private channel. Do not include any details in that first message. A useful report includes: - the LearnHarness version and your OS; - what an attacker can do and what they need first (a web page the agent reads, a file the learner opens, network access, and so on); - steps to reproduce, or a proof of concept; - any fix or mitigation you have in mind. ## Security contacts - [@amosblomqvist](https://github.com/amosblomqvist) - [@IsaacZhangg](https://github.com/IsaacZhangg) Use the private reporting process above for vulnerability reports. Do not report security details through public mentions. ## What to expect These are targets, not guarantees: | Step | Target | | --- | --- | | Acknowledge your report | within 7 days | | Initial assessment (confirmed or not, severity) | within 14 days | | Fix released and advisory published | within 90 days of the report | The fix is developed privately in a GitHub security advisory. We will keep you updated, agree on a disclosure date with you, and credit you in the published advisory unless you prefer to stay anonymous. A CVE can be requested through GitHub when one is warranted. ## Supported versions Only the latest release receives security fixes. The desktop app updates itself, so fixes reach users through a new release. If you run from source, update to the latest `main` or the latest release tag. ## Scope In scope: - the desktop app (Electron shell, window, and IPC); - the local API server, whether started by the desktop app or run from source; - the agent and its tools, including prompt injection from web pages, PDFs, or other content that leads to reading, changing, or sending your data; - rendered content such as demos, figures, and markdown; - the update and release pipeline. Out of scope: - attacks that already require running code as your OS user; - bugs in third-party model or service providers themselves (report those to the provider); - automated scanner output without a demonstrated impact on LearnHarness.