The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently spawns the curl binary using spawnSync to handle multipart and binary data for the Cloudflare API.
This pattern is found in scripts/workers.ts for deploying worker scripts and scripts/r2-storage.ts for uploading and downloading objects.
While the scripts include basic validation (e.g., regex for worker names), the use of external binaries for network operations is a sensitive capability.
[DATA_EXFILTRATION]: The skill reads sensitive local files to perform its primary functions.
It accesses the .env file in the project root to retrieve the CLOUDFLARE_API_KEY for authentication.
It reads local application code and assets to deploy them to Cloudflare Workers and R2 buckets.
All network communication is directed to Cloudflare's official API (api.cloudflare.com), which is a well-known and trusted service.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to how it handles data retrieved from remote storage.
Ingestion points: Data is fetched from Cloudflare KV namespaces in scripts/kv-storage.ts and R2 buckets in scripts/r2-storage.ts and then displayed or saved locally.
Boundary markers: The skill does not provide any boundary markers or instructions to the agent to disregard potential commands embedded in the fetched data.
Capability inventory: The skill possesses file-write, network-send (fetch/curl), and subprocess-execution capabilities, making an injection potentially impactful.
Sanitization: There is no sanitization or validation of the content fetched from remote storage before it is placed into the agent's context.

Cloudflare Manager