application-performance-performance-optimization
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill interpolates user-controlled input ($ARGUMENTS) into instructions for various high-capability sub-agents without implementing boundary markers or sanitization. This presents an indirect prompt injection surface.\n
- Ingestion points: User input from $ARGUMENTS is used across all profiling, optimization, and testing phases (e.g., Steps 1, 4, 7, 10).\n
- Boundary markers: There are no delimiters or 'ignore-instructions' guards surrounding the interpolated input.\n
- Capability inventory: Sub-agents (e.g., performance-engineer, database-optimizer, test-automator) are tasked with modifying database configurations, backend code, and CI/CD pipelines (GitHub Actions).\n
- Sanitization: The workflow does not include steps to validate or sanitize the input before it reaches the sub-agents.\n- [DATA_EXFILTRATION]: The skill involves access to sensitive data types for profiling and bottleneck identification, which is an expected part of its performance optimization functionality.\n
- Evidence: The instructions include generating heap dumps for memory analysis, tracing I/O operations, and analyzing database slow query logs.\n
- Context: These actions are scoped to performance profiling and are accompanied by instructions to use established APM tools like DataDog or New Relic.
Audit Metadata