The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill possesses a significant attack surface for indirect prompt injection.
Ingestion points: Processes JSON, YAML, and TOML data from external sources such as API responses, Kubernetes manifests, and Docker Compose files (SKILL.md).
Boundary markers: No boundary markers or instructions to ignore embedded commands are present.
Capability inventory: Utilizes Bash to execute jq and yq. Notably includes file-write capabilities via yq -i (in-place modification) as seen in SKILL.md.
Sanitization: No evidence of data sanitization or validation of the processed structured data.
Command Execution (MEDIUM): The skill facilitates the execution of CLI tools (jq, yq) on the host system. While these are specialized tools, yq -i allows for modification of local configuration files which could be used to weaken system security or persist changes.

data-processing