pp-airflow-admin

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill downloads and installs software from external sources using go install and npx. The target resources (e.g., github.com/mvanhorn/printing-press-library) are associated with the skill vendor 'mvanhorn'.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill provides a --deliver webhook:<url> feature that allows sending command results to arbitrary external URLs. This functionality could be used to exfiltrate sensitive Airflow metadata such as connection secrets or variables.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Apache Airflow APIs, creating a surface for indirect prompt injection. Ingestion points: data enters the context through dags list, dag-runs list-task-instances, and sync commands in SKILL.md. Boundary markers: None are present to delimit external content or warn the agent. Capability inventory: The skill has access to the Read Bash tool for command execution and network delivery via webhooks. Sanitization: No escaping or validation of external content is mentioned.
  • [COMMAND_EXECUTION]: The skill executes a custom binary airflow-admin-pp-cli and setup tools like go and npx using the Read Bash tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 08:25 AM
Security Audit — agent-trust-hub — pp-airflow-admin