analytics-tracking
Pass
Audited by Gen Agent Trust Hub on Feb 19, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection (LOW): The skill's primary instruction set in
SKILL.mddirects the agent to read.claude/product-marketing-context.mdbefore proceeding. This allows external, unvalidated content to influence the agent's logic and the resulting tracking plans or code snippets it generates.\n - Ingestion points:
.claude/product-marketing-context.md(referenced in the 'Initial Assessment' section ofSKILL.md).\n - Boundary markers: Absent; the skill does not provide delimiters or warnings to ignore potential instructions within the context file.\n
- Capability inventory: The skill generates executable JavaScript code for analytics implementation (GA4
gtag, GTMdataLayer, and Facebook Pixel) and designs strategic measurement plans.\n - Sanitization: Absent; the agent is instructed to use the file context directly to inform its behavior without any validation or filtering steps.
Audit Metadata