Skip to content

Security: RushObservability/sre-agent

SECURITY.md

Security Policy

Reporting a Vulnerability

Please do not report security vulnerabilities through public GitHub issues, pull requests, or discussions.

Report privately through GitHub Security Advisories:

  1. Go to the Security tab
  2. Click Report a vulnerability

This opens a private channel visible only to the maintainers. If you cannot use GitHub Security Advisories, open a regular issue asking for a private contact channel -- without including any vulnerability details.

What to include

  • The type of issue (authentication bypass, injection, privilege escalation, ...)
  • Affected source paths, and the commit or version you tested
  • Steps to reproduce, including any configuration required
  • Impact: what an attacker can achieve

What to expect

  • Acknowledgement within 5 business days.
  • Initial assessment, including whether we accept the report and a severity judgement, within 10 business days.
  • Progress updates at least every 10 business days until resolution.
  • Credit in the advisory and release notes, unless you prefer to remain anonymous.

We ask that you give us a reasonable opportunity to ship a fix before public disclosure. We aim to release fixes for confirmed high and critical issues within 90 days of the report.

Supported Versions

This project is pre-1.0 and moves fast. Security fixes land on main and in the most recent release only; there are no backports to earlier tags.

Version Supported
main and latest release Yes
Anything older No

Scope

This agent executes investigations against live infrastructure and holds credentials for query-api, Kubernetes, and LLM providers. It does not receive ClickHouse credentials. Reports touching the following are especially valuable:

  • Authentication and authorization on the HTTP API, including JWT handling
  • Command, query, or prompt injection reaching Kubernetes or query-api tools
  • Leakage of credentials or customer telemetry into logs, traces or LLM prompts
  • Sandbox escape from any tool the agent can invoke

Out of scope: findings against your own deployment's configuration (network exposure, credentials you supplied, IAM policy), volumetric denial of service, and reports generated by automated scanners without a demonstrated impact.

Our Automated Security Practices

  • Dependabot version updates for Cargo, Docker and GitHub Actions, plus Dependabot alerts against the GitHub Advisory Database (a different advisory set from RustSec -- neither covers the other)
  • cargo audit runs in CI and fails the build on RustSec vulnerability advisories
  • CodeQL static analysis on every push and pull request
  • Dependency review blocks pull requests introducing vulnerable dependencies
  • All GitHub Actions are pinned to full commit SHAs
  • Workflow tokens are scoped to least privilege

There aren't any published security advisories