liveness_primer is pre-1.0 and under active development. Only the latest release on PyPI and the current main branch receive security fixes; there are no backports to earlier versions.

Report privately through GitHub's private vulnerability reporting (the "Report a vulnerability" button on the repository's Security tab). This opens a draft advisory visible only to you and the maintainer.

Please do not open a public issue for a suspected vulnerability.

Include what you have: the affected version or commit, the command line and corpus entry that triggers it, and what an attacker gains. A proof of concept helps but is not required to file.

This is currently a single maintainer project. We aim to acknowledge valid reports within three days and initial triage within about a a week. Critical issues may be handled on an accelerated timeline.

There is no bug bounty.

liveness_primer deliberately clones third-party repositories, installs detector revisions from arbitrary refs, and executes those detectors as subprocesses. Running it means running other people's code. Reports about that design are in scope only where the tool fails to hold a boundary it claims to hold — for example:

• Detector or corpus code escaping the network isolation the runner asserts (liveness_primer/isolation.py), or the run being reported as isolated when it was not.
• Path traversal in liveness_primer's own cache management that causes a checkout or detector environment to be created outside the configured cache root (liveness_primer/corpus.py, envcache.py). This is not a general filesystem sandbox: detector build hooks and subprocesses retain the current user's access to host-visible paths.
• Corpus or report input causing code execution during parsing — the corpus is loaded with PyYAML's safe loader and validated by pydantic, so anything that reaches eval/exec/arbitrary constructors is a bug.
• Secrets (tokens, environment contents) leaking into reports, logs, or subprocess environments.

Out of scope:

• Vulnerabilities in the detectors themselves (vulture, skylos) or in corpus projects — report those to their maintainers.
• Wrong, missing, or noisy detector findings. That is a correctness bug; open a normal issue.
• Running detector refs you do not trust without external filesystem and resource isolation. The built-in isolation removes network access and credentials but does not restrict filesystem access; run such comparisons in a disposable container.