flow-management

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted project metadata.
  • Ingestion points: The skill ingests untrusted data from the Dataiku environment via project.list_datasets() and project.list_recipes() in SKILL.md and references/build-strategies.md.
  • Boundary markers: There are no boundary markers or explicit instructions to ignore embedded commands within the names of datasets or recipes.
  • Capability inventory: The skill possesses significant capabilities, including triggering project jobs and builds via recipe.run(), job.start(), and job.start_and_wait().
  • Sanitization: No sanitization or validation is applied to object names retrieved from the API before they are processed or potentially interpolated into prompts.
  • [COMMAND_EXECUTION]: The skill uses the Dataiku Python API to trigger execution of recipes and jobs (recipe.run(), project.new_job().start()). While these are standard for flow orchestration, they represent the primary execution capabilities available to an agent utilizing this skill.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 01:37 AM