grove-auth-integration
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input such as Project Name, Client ID, and Site URL in Step 1. This data is then directly interpolated into shell commands and SQL queries in Step 2c and Step 3 without explicit sanitization or validation instructions for the agent. This creates an attack surface where a malicious user could provide input designed to execute arbitrary commands or manipulate the D1 database when the agent runs the generated wrangler commands.\n
- Ingestion points: SKILL.md Step 1 (Project metadata collected from user).\n
- Boundary markers: Absent. Variables are placed directly into template strings.\n
- Capability inventory: shell command execution (wrangler), database execution (wrangler d1), and file system modification (writing SvelteKit routes).\n
- Sanitization: Absent.\n- [COMMAND_EXECUTION]: The skill guides the agent to execute administrative shell commands including 'wrangler d1 execute' for database manipulation and 'wrangler pages secret put' for secret management. While these are necessary for the skill's purpose, the lack of input validation on the variables used in these commands presents a security risk.
Audit Metadata