development-assistant
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns and templates for creating 'Analyzers' that process untrusted patent content (
content: str). This introduces a vulnerability surface where malicious instructions embedded in patent data could influence the agent's behavior during analysis. - Ingestion points: The
analyze(self, content: str)method in theBaseAnalyzertemplate (SKILL.md) accepts raw strings for processing. - Boundary markers: The provided templates do not include delimiters or instructions to ignore embedded commands in the input data.
- Capability inventory: The system includes capabilities for BigQuery interaction (
mcp_server/bigquery_search.py), performance monitoring, and creating new tools (mcp_server/server.py). - Sanitization: The development workflow and code templates do not explicitly mention sanitization or validation of the processed text content beyond Pydantic input model validation for tool parameters.
Audit Metadata