The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npm create cloudflare which pulls templates from github.com/cloudflare/ai, a well-known repository from a trusted service provider. This is a standard and expected part of the project initialization workflow.
[PROMPT_INJECTION]: The query_db tool implementation in SKILL.md creates an indirect prompt injection surface by accepting a raw sql string parameter and executing it directly against the database binding (env.DB.prepare(sql).all()). This lacks the recommended evidence chain protections: 1. Ingestion point: sql parameter in SKILL.md. 2. Boundary markers: No explicit instructions to ignore embedded commands are present in the code or prompt context. 3. Capability inventory: The tool has the capability to perform arbitrary SQL execution (read/write) against the Cloudflare D1 database binding (env.DB). 4. Sanitization: Absence of parameterization, input validation, or escaping in the provided code example.
[COMMAND_EXECUTION]: The skill instructs users to perform development tasks using tools like wrangler and npm, including deploying code to production environments (wrangler deploy) and running local testing environments. These commands are standard and necessary for the advertised functionality of the skill.

building-mcp-server-on-cloudflare