We treat security reports as the highest-priority issue category. This document is the contract: how we want to be told about vulnerabilities, what we promise back, and how the CLI handles your credentials so you can audit its trust boundary before you adopt it.
Please do not open public GitHub issues for security bugs. Use one of the private channels below.
- Preferred: GitHub Security Advisories — encrypted, tied to the repo, lets us collaborate on a fix and a CVE.
- Email:
security@ararahq.com— for issues you'd rather not file via GitHub. Subject line should start with[arara-cli security]. PGP key on request.
Include in your report (best effort, anything you can give helps):
- A short description of the issue and its impact.
- Steps to reproduce, plus the CLI version (
arara --version) and OS. - Whether you've already disclosed this anywhere else.
- How you'd like to be credited in the advisory (or "anonymous").
| When | What |
|---|---|
| Within 24 business hours | Acknowledge receipt and assign an owner. |
| Within 5 business days | Confirm whether it's reproducible and give a severity estimate. |
| Before public disclosure | Coordinate a fix window with you; default 90 days from confirmation, faster if there's an active exploit. |
| After fix lands | Credit you in the advisory unless you ask us not to; tag the CVE if one was assigned. |
We don't have a paid bug bounty yet — but contributions to the security of the CLI are taken seriously and credited publicly.
We patch security fixes against the latest minor release. Older minor releases get fixes on a best-effort basis only — upgrade is the canonical path.
| Version | Supported |
|---|---|
Latest minor (currently 0.x) |
✅ |
| Anything older |
arara upgrade --install updates in place, verifies the SHA256 of the
new release, and preserves your config.
The CLI handles two classes of credential: API keys (ara_live_… /
ara_test_…) and OAuth tokens (returned by the Device Authorization
flow). Both are treated as bearer secrets.
- OS keyring first. macOS Keychain, Windows Credential Manager, or
Linux Secret Service (libsecret / kwallet). The CLI calls
github.com/zalando/go-keyring; we never see plaintext after the first write. ~/.config/arara/config.tomlfallback. If the keyring is not available (CI containers, headless servers, broken libsecret), the key is written to the user's config file with permissions0600. The file path is logged so you know it happened — there is no silent fallback.- No environment-variable leak. The CLI does not read
ARARA_API_KEYfromprocess.envautomatically. You have to opt in witharara login --key "$ARARA_API_KEY".
- The
Clientholds the active token for the lifetime of the process. - Subprocess and HTTP middleware redact tokens before logging
(see
internal/api/redact.go). The redactor strips by prefix (ara_live_,ara_test_,Bearer …,wh_cli_,whsec_…) and by field name (token,secret,apiKey,authorization,password).
- All API requests are HTTPS-only. There is no
--insecureflag. - Bearer tokens go in the
Authorizationheader, never in the URL or query string.
These are documented so they're not surprises:
- Config file fallback is plaintext. Use
chmod 600 ~/.config/arara/config.tomlif you don't trust your OS keyring (default permissions are already0600, but worth confirming on shared boxes). - Curl-pipe-shell install (
curl … | sh) trusts GitHub and the TLS chain. SHA256 checksums are verified against the release'schecksums.txtafter download. To inspect the script first:curl -fsSL https://raw.githubusercontent.com/ararahq/cli/main/install.sh > install.sh less install.sh sh ./install.sh - OS keyring on Linux depends on the desktop environment. Headless
servers will hit the file fallback unless you have
passor another Secret Service provider configured.
These are not vulnerabilities we'll act on:
- Social engineering against AraraHQ employees.
- Denial of service against
api.ararahq.comfrom your own account. - Issues that require physical access to an unlocked workstation that already has a logged-in CLI session.
- Self-XSS via terminal escape sequences in your own output. (We sanitize remote inputs; we trust the local environment.)
Researchers who report valid issues to us in good faith: