argus is designed to be safe to run on a developer or security engineer's own machine:

• It is read-only on scan. It never deletes, kills, quarantines, chmods, or installs anything without explicit, per-run consent.
• It never executes a suspect or audited artifact. Suspect package versions are read from the filesystem; a fetched git URL is shallow-cloned with hooks and transports disabled and nothing in it is run.
• It never phones home by default. The only network actions are opt-in and announced.
• It redacts matched content to file:line, so it does not spill secrets into its report.

The only mutation argus will perform is a consent-gated install of a named security engine (Trivy, pinned and checksum-verified; or ClamAV via the OS package manager). CI, MDM, and --no-install (or the ARGUS_NO_INSTALL environment variable) disable all prompts and installs.

If you find a security issue in argus itself (for example a way to make it execute a scanned artifact, escape the read-only contract, or exfiltrate data), please report it privately:

• Open a GitHub security advisory on this repository (Security tab, "Report a vulnerability"), or
• Contact the maintainers listed in .github/CODEOWNERS.

Please do not open a public issue for a security report. We aim to acknowledge reports promptly and will coordinate a fix and disclosure timeline with you.

The main branch is the supported line. Security fixes are applied there and documented in CHANGELOG.md.

argus is public by design. Its detection logic, IOC dataset, and audit rules are intended to be read; outsiders can only use the tool to scan their own machines. Reports about the detection content (false positives, missed indicators, rule bypasses) are welcome as normal issues or pull requests; see CONTRIBUTING.md.