lovable-project
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script
plugins/lovable/scripts/detect-architecture.pyrelative to the project root to detect the tech stack (Vite vs TanStack). This is a standard project-specific helper command. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon content from external repository files, creating a potential surface for indirect instructions to influence agent behavior.
- Ingestion points: Processes
.lovable-agent/config.json,.lovable-agent/context.md,AGENTS.md, and legacyCLAUDE.mdfiles from the repository. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified for these files.
- Capability inventory: The agent is granted capabilities to execute local Python scripts and read/modify project files based on the ingested context.
- Sanitization: The skill provides guidelines for secret safety (e.g., recording names only, not values), but does not implement validation or escaping for natural language instructions embedded within the ingested files.
Audit Metadata