deploying-cloudflare-web-application
Pass
Audited by Gen Agent Trust Hub on Aug 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository configuration files, including
package.json,pnpm-workspace.yaml, and various framework-specific configuration files (e.g.,vite.config.ts,next.config.js), as detailed inreferences/discovering-application.md. This data is used to determine and execute shell commands such as<build-command>andnpx wrangler deploywithin the environment. - Ingestion points: Project configuration files (
package.json,wrangler.toml, workspace configs, framework configs) processed during the discovery phase inreferences/discovering-application.md. - Boundary markers: The instructions do not define explicit boundary markers or safety delimiters when processing content from these external files to differentiate between data and instructions.
- Capability inventory: The skill instructs the agent to execute shell commands (
bash,ls,npx) and write to the local filesystem (wrangler.toml,package.json) based on the discovered configuration. - Sanitization: There is no evidence of sanitization or validation of the extracted commands or configuration values before they are used in shell execution, relying on the agent's interpretation.
Audit Metadata