Devicetype Library CVE-2026-55251

Exploitation requires that the repository's .github/workflows/validation.yml run on the pull_request event and execute PR-supplied content before maintainer review - specifically via contributor edits to requirements.txt (pip install), .pre-commit-hooks-config.yaml / .pre-commit-yamlfmt-config.yaml (pre-commit hook execution), or .gitmodules (submodule checkout). The attacker only needs the ability to open a pull request (no repository write access), and no user interaction from a maintainer is needed because CI runs automatically. Exploitation is LIMITED to versions prior to commit f41fc1e, and the achievable impact is constrained because fork PRs run with a read-only token and no secrets; the patch adds a validate-inputs gate confining untrusted authors to the device-types/, module-types/, rack-types/, elevation-images/, and module-images/ contribution directories unless the author is trusted or a maintainer applies the ci-approved label.

This is a medium-severity CI/CD supply-chain weakness (CWE-94) in the NetBox devicetype-library repository's GitHub Actions workflow, scored CVSS 6.5 (AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:N). The network vector and no-privilege, no-interaction metrics are accurate in the GitHub sense: any person with a GitHub account can open a pull request, and the vulnerable validation.yml ran PR-supplied code automatically before maintainer review. No EPSS score, CISA KEV listing, or POC was supplied in the intelligence, so there is no public exploit identified at time of analysis and no signal of active exploitation - this is a disclosed-and-patched issue rather than a fire. Real-world risk is tempered by two factors visible in the patch commit's own commentary: fork PRs run with a read-only token and no secrets, which is precisely why impact is capped at C:L/I:L/A:N rather than a high-confidentiality or integrity compromise. The practical worst case is arbitrary code execution on an ephemeral CI runner with limited privileges (e.g., cryptomining, abusing runner compute, or probing for misconfigured secrets), not theft of repository secrets or production compromise. Net: a legitimate issue worth patching that is already fixed, but not a high-urgency widespread threat; treat it as a medium priority confined to the single affected repository's CI.

An attacker with an ordinary GitHub account forks the repository and opens a pull request that edits requirements.txt, a pre-commit hooks config, or .gitmodules to point at or install attacker-controlled code. When validation.yml triggers on the pull_request event, the CI runner installs/executes that code before any maintainer reviews the PR, giving the attacker code execution on the runner. No public exploit code is identified at time of analysis, and the runner executes with a read-only token and no secrets, limiting the blast radius.