write-endpoints

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [External Data Handling]: The skill illustrates how to build endpoints that process untrusted data from HTTP requests. While this creates a potential surface for input-based vulnerabilities, the skill proactively addresses these considerations by providing instructions and utilities for strict schema validation and SQL injection prevention.
  • Ingestion points: The OpenAPIRoute and D1Endpoint classes define methods to ingest body, query, and header data from incoming requests in SKILL.md.
  • Boundary markers: The documentation relies on Zod schema definitions to strictly validate and constrain incoming data types.
  • Capability inventory: The skill templates perform database operations using Cloudflare D1 integration.
  • Sanitization: The skill provides specific methods for validating SQL identifiers and building safe filters (validateTableName, validateColumnName, buildSafeFilters) to prevent injection attacks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 04:07 PM
Security Audit — agent-trust-hub — write-endpoints