debugging-dags

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE] (SAFE): No malicious instructions, prompt injections, or unauthorized data access patterns were identified. The skill's behavior is consistent with its stated purpose of diagnostic support.\n- [NO_CODE] (SAFE): The skill is comprised entirely of markdown instructions and does not include any Python scripts, Node.js packages, or binary files.\n- [Indirect Prompt Injection] (SAFE): Evaluation of the potential for malicious instructions within processed DAG logs or source code. \n
  • Ingestion points: The skill accesses external data through get_failed_runs, analyse_dag_latest_run, get_dag_run_detail, get_dag_source_code, and go_to_dag_log_view (SKILL.md).\n
  • Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands in the logs or code.\n
  • Capability inventory: The agent has the ability to trigger actions via trigger_dag_run (SKILL.md).\n
  • Sanitization: No explicit data sanitization or validation of the ingested Airflow content is defined in the skill markdown.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:22 PM