Grav CMS EUVD-2026-81439

Requires an authenticated account with page-edit rights AND a page with Twig processing enabled on its content - both are prerequisites named in the description. Grav 2.0.0-rc.1 through 2.0.21 only; Grav 1.7 is NOT affected because it ships no Twig content sandbox, and 2.0.22 fixes it by registering the filters with Twig's needs_is_sandboxed flag. The exploit uses the dump/serialize filters (print_r, vardump, json_encode, yaml_encode, string). Limiting factors: attacker must first obtain editor-level access, so this is a privilege-to-secrets escalation rather than an unauthenticated exposure; no user interaction from a victim is needed once the malicious page renders.

This is a genuine but moderate-priority information-disclosure flaw, not a critical remote-takeover. The vendor CVSS 4.0 vector (8.7) sets PR:N, but the description explicitly requires 'an authenticated user with page-edit rights' - a direct contradiction that materially lowers real risk. The vendor score should read PR:L; my reassessment lands around CVSS 3.1 6.5. Impact is confidentiality-only (VC:H, VI:N, VA:N) with no code execution or data tampering. There is no EPSS score, no CISA KEV listing, and no public POC in the provided data - so no evidence of active or automated exploitation. The realistic threat is a low-privilege editor (or a compromised editor account) escalating access to secrets rather than an unauthenticated internet-wide attack. CONFLICT TO VERIFY: CVSS PR:N vs. description's 'authenticated user with page-edit rights' - treat this as an authenticated flaw, not unauthenticated.

An attacker who already holds (or has phished/compromised) a Grav account with page-edit rights creates or edits a page with Twig processing enabled and inserts {{ config|print_r }} in the content. When the page renders, print_r reflects the real Config object behind the SandboxConfig facade and bypasses path redaction, dumping the entire merged configuration including SMTP credentials, API tokens, webhook secrets and cache backend passwords. No public exploit code is identified in the provided data, though the technique is trivially reproducible from the advisory.