audit-db-schema
Fail
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent to read sensitive configuration files, specifically '.env' and '.env.local', to discover database project IDs and connection strings.
- [COMMAND_EXECUTION]: The skill uses the 'execute_sql' tool to run arbitrary, dynamically generated SQL queries against the database schema. This capability provides broad access to the database and can be exploited to perform unauthorized operations if the agent is misled.
- [EXTERNAL_DOWNLOADS]: The skill performs automated research by conducting web searches and scraping external sites via 'firecrawl_search' and 'firecrawl_scrape' to gather database patterns and documentation.
- [PROMPT_INJECTION]: The inclusion of untrusted content from the web (via Firecrawl scraping) creates a surface for indirect prompt injection. Malicious instructions placed on external websites could potentially influence the agent's behavior during the audit.
- Ingestion points: Results from 'firecrawl_scrape' in SKILL.md.
- Boundary markers: None identified; external content is processed without isolation or 'ignore' instructions.
- Capability inventory: The skill has permission to execute arbitrary SQL, read schema metadata, and list all tables across various database providers.
- Sanitization: No sanitization or validation of the scraped web content is described before it is used to generate the audit report.
Recommendations
- AI detected serious security threats
Audit Metadata