typescript-mcp
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of MCP servers that ingest untrusted data from LLMs or external clients via HTTP POST requests.
- Ingestion points: Data enters through
c.req.json()inSKILL.md,templates/basic-mcp-server.ts, and other template files where the Hono app processes the incoming JSON-RPC payload. - Boundary markers: The skill strongly encourages the use of Zod for schema validation in all tool definitions (documented in
SKILL.mdandtemplates/tool-server.ts), which acts as a primary defense against malformed or malicious inputs by enforcing strict data types and structures. - Capability inventory: Depending on the implementation chosen by the user, the generated servers have capabilities for D1 database queries, KV store operations, R2 storage access, and external network requests via
fetch. - Sanitization: The skill includes explicit guidance in
references/tool-patterns.mdto use parameterized queries to prevent SQL injection and to sanitize error messages before returning them to the client to avoid leaking sensitive information.
Audit Metadata