generating-and-analyzing-sboms

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow in scripts/agent.py uses Grype by invoking grype sbom:{sbom_path} and then parses the resulting Grype JSON via json.loads (lines 54-65 and 68-75), meaning outsider-authored text can be ingested if an attacker can cause a queued/posted SBOM file path or image/SBOM content to be supplied to the --sbom/--sbom-out inputs.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The SKILL includes prerequisite install commands that fetch and execute remote install scripts (curl -sSfL https://raw.githubusercontent.com/anchore/syft/main/install.sh | sh and curl -sSfL https://raw.githubusercontent.com/anchore/grype/main/install.sh | sh) and a go install that pulls/builds remote code (go install github.com/sigstore/cosign/v2/cmd/cosign@latest), which directly execute external code and are presented as required setup steps.