Report suspected vulnerabilities privately through GitHub Security Advisories: Report a vulnerability.

Please do not open a public issue or pull request for a security problem.

Include the command you ran, the version (splunk --version), and what you observed. You will get an acknowledgement within 7 days and a status update at least every 14 days until the report is resolved.

The latest release on main receives security fixes. Older versions do not.

• Secret credentials come from the environment or a configuration profile, never from a command-line flag. Flags are recorded in shell history and are visible to any other user in the process list, so no option accepts a password, token, or session key. auth login accepts --username, which is not a secret; it reads the password from $SPLUNK_PASSWORD or a no-echo prompt.
• Credentials are never written to disk by this tool. It reads them, uses them for the current invocation, and forgets them.
• Secret-bearing response fields are replaced before data leaves an operation. Splunk returns secrets inside ordinary reads — an HTTP Event Collector input carries its own token, a server setting carries pass4SymmKey. Whether a field is secret is decided by its name, in one place (core/redact.py), and applied to every named read command on both backends. The key survives as <redacted> so you can still see the field exists, and a field Splunk adds later is covered the day it appears.
• api get is the one exception, by design. It is a raw escape hatch that returns the endpoint's body verbatim, so it can show a stored secret. It stays verbatim on purpose: redacting it would corrupt the round-trip that makes an escape hatch useful. Prefer the named command when one exists.
• Three commands return a secret on purpose, because minting one is what they do: auth login prints the session key it created, hec rotate prints the token it regenerated, and hec-token create prints the token it just created — you cannot configure a sender without it. Those values go to standard output only, never to the audit log, which records the action and the object name. Every other named command redacts.
• TLS verification is on by default. SPLUNK_CA_BUNDLE points at your own certificate authority. SPLUNK_VERIFY=false disables verification entirely and exists only for laboratory use.

Reads cannot modify a server. Every write passes through one gate:

• --dry-run prints the exact request and sends nothing.
• On a terminal, the write asks for confirmation.
• Without a terminal, the write requires --yes and otherwise stops. It never waits on a prompt nobody can see.
• Every applied write is appended to a local audit log.

An operation the target does not support fails with a typed error. The tool never falls back to guessing an endpoint.

• Two runtime dependencies: click and httpx. Every added package is both maintenance burden and attack surface, so the list does not grow without a written case that the standard library cannot do the job.
• Every GitHub Actions reference is pinned to a full commit SHA — third-party actions and reusable workflows alike. A branch or tag can be repointed at new code after review; a commit cannot.
• Workflows declare least-privilege permissions: blocks.
• Every pull request runs CodeQL (GitHub code scanning default setup, so there is no workflow file for it), zizmor workflow-security analysis, and dependency review that fails on moderate or higher severity.
• OpenSSF Scorecard audits these practices weekly and publishes the result.