The Agent Skills Directory

Remote Code Execution (CRITICAL): The skill recommends installing the Syft tool via curl -sSfL https://raw.githubusercontent.com/anchore/syft/main/install.sh | sh. This pattern downloads a script from the internet and pipes it directly into a shell for execution. Because the 'anchore' GitHub organization is not within the defined trusted scope, this is classified as an untrusted remote code execution finding.
Indirect Prompt Injection (HIGH): The skill is designed to ingest and process untrusted data from external directories (syft dir:<target-path>) and container images (syft <image>:<tag>).
Ingestion points: SKILL.md instructions for scanning local paths and remote container images.
Boundary markers: None detected. The skill does not use delimiters or instructions to ignore embedded malicious content within the files it scans.
Capability inventory: The skill executes system-level commands and generates summaries that could influence downstream security decisions or agent actions.
Sanitization: No evidence of sanitization or validation of the scanned data or its output metadata exists.
Command Execution (HIGH): The skill facilitates the execution of arbitrary system commands using the syft binary on user-provided paths or images, which can be exploited if the inputs are not strictly controlled.

sbom-syft