shift-right-testing

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting and acting upon untrusted external data from production systems.
  • Ingestion points: Production logs are ingested in 'evals/shift-right-testing.yaml' (tc003) and structured incident data is processed in the 'Incident Replay' task in 'SKILL.md'.
  • Boundary markers: Absent; there are no instructions or delimiters defined to isolate external data from the agent's core instructions.
  • Capability inventory: The skill can coordinate a mesh of agents via 'FleetManager' and execute automated tasks including test generation and chaos experiment injection.
  • Sanitization: The instructions do not define any sanitization or validation logic for the input logs or incident descriptions.
  • [COMMAND_EXECUTION]: The skill demonstrates dynamic task orchestration and automated test generation capabilities.
  • Evidence: The 'Incident Replay' task in 'SKILL.md' is explicitly designed to 'generateTests' and 'addToRegression' based on the conditions of a production incident.
  • Evidence: Uses 'FleetManager.coordinate' to dynamically manage a mesh topology of specialized agents ('qe-chaos-engineer', 'qe-performance-tester', etc.) to execute production-level quality engineering workflows.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 09:39 AM