mcp-light-generator
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill generates instructions and Python code templates that use shell commands (e.g.,
npx) to wrap existing MCP servers. This is an intended and standard implementation of theFastMCPproxy pattern. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes tool descriptions from external MCP servers to generate new agent skills. This represents a potential surface for indirect prompt injection if the source tool metadata contains malicious instructions; however, this behavior is inherent to the skill's primary function of metadata optimization.
- Ingestion points: Tool descriptions fetched via
ToolSearchfrom specified MCP servers. - Boundary markers: None explicitly implemented in the generated skill templates.
- Capability inventory: File writing (server.py, SKILL.md, pyproject.toml) and generating command-line installation instructions.
- Sanitization: The skill applies summarization and analysis to the source descriptions, which may alter but does not explicitly sanitize injection attempts.
Audit Metadata