optimizing-performance
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists of instructional documentation and code snippets representing industry-standard best practices for performance optimization. No evidence of prompt injection, malicious commands, or unauthorized data access was found.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a standard attack surface for performance analysis by processing external inputs.
- Ingestion points: The skill utilizes local profiling log files (
isolate*.log,profile.stats) and external web pages analyzed via the Lighthouse CLI as data sources. - Boundary markers: The instructions do not specify specific delimiters or boundary markers to separate external data from agent instructions.
- Capability inventory: The skill involves command execution for profiling tools (
node,python,lighthouse) and file system redirection to save logs. - Sanitization: No explicit sanitization or filtering of log content or external web page data is described.
- Contextual Assessment: While these components define a potential ingestion surface for indirect prompt injection, this behavior is essential for the skill's primary purpose of performance auditing. The risk is consistent with standard developer tooling and is assessed as safe in this context.
Audit Metadata