The Sightline maintainer takes security bugs seriously. We appreciate your efforts to responsibly disclose your findings, and will make every effort to acknowledge your contributions.

Do NOT open a public GitHub issue for security vulnerabilities.

Instead, please report vulnerabilities privately:

1. Email: See GitHub Security Advisories (preferred)
2. Subject line: [SECURITY] Sightline — <short description>
3. Include in your report:
   • Description of the vulnerability and its potential impact
   • Steps to reproduce (proof of concept if possible)
   • Affected versions (commit hash or release tag)
   • Any suggested mitigations or fixes

You should receive an acknowledgment within 72 hours. If you do not receive an acknowledgment within that time, please follow up by email.

• We follow coordinated disclosure. We will not disclose a vulnerability publicly until a fix is available, unless the reporter agrees otherwise.
• We will credit the reporter in the release notes / security advisory, unless they prefer to remain anonymous.
• We request that reporters do not publicly disclose the vulnerability until a fix is available, or until 90 days have passed since the initial report (whichever comes first).

Sightline is pre-1.0 software. Security fixes are applied only to the latest release on the main branch. No backports to older versions are provided.

Sightline has been hardened with the following measures (as of the current release):

• Authentication: Firebase Auth + role-based access control (RBAC) (require_auth, require_admin, require_role, optional_auth).
• Dev mode: Local-only bypass (DEV_AUTH_BYPASS=true + SERVER_HOST=127.0.0.1). Blocked on 0.0.0.0 and public IPs.
• Rate limiting: Per-IP request limits + role-aware message limits.
• Path traversal protection: Path.is_relative_to() containment on all file-serving endpoints.
• XSS protection: sanitizeHtml(md(...)) for chat output, esc() with single-quote escape for all dynamic HTML.
• CSP: object-src 'none'; base-uri 'self'; form-action 'self' plus HSTS, X-Frame-Options, X-Content-Type-Options, Referrer-Policy.
• SQL injection: SQL query tool is read-only (SELECT only) with a 5-second timeout.
• Upload security: %PDF- magic byte check + secure_filename.
• Error sanitization: Generic error messages to clients, details in server logs only.
• Stream auth: SITREP stream uses single-use nonce (not JWT in URL), bound to (uid, job_id) with 5-minute TTL.
• Test coverage: 250 tests, including security tests (path traversal, dev mode safety, stream nonce, SQL read-only, PII redaction, role-based tool filtering).

• Security vulnerabilities in the Sightline application code (Python, JavaScript).
• Authentication / authorization bypasses.
• Server-side request forgery, injection (SQL, command, XSS).
• Path traversal, file disclosure.
• Sensitive data exposure (API keys, credentials, PII).
• Security-relevant misconfigurations (CSP, CORS, cookies).

• Vulnerabilities in third-party dependencies (report to the upstream project).
• Vulnerabilities in the underlying infrastructure (OS, Docker, Node.js, Python runtime).
• Social engineering, phishing.
• Denial of service via volume (we have rate limiting; DoS resilience is best-effort, not a security guarantee).
• Theoretical issues without a working proof of concept.
• Issues requiring physical access to the server or user's machine.
• Bugs in the LLM agent's tool-calling behavior that do not have a security impact.

Sightline is a pre-revenue open-source project and does not currently offer a monetary bug bounty. We genuinely appreciate responsible disclosure and will credit reporters in release notes.

(If you would like to sponsor a bug bounty fund, please contact the maintainer.)

• Security reports: GitHub Security Advisories (preferred)
• General questions: open a GitHub Discussion

Sightline Security Policy v1.0