The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/probe-ci-token.sh is vulnerable to command injection via its command-line arguments (specifically the account ID and zone name parameters). These inputs are unsafely interpolated into a generated shell script that is intended to run within a GitHub Actions environment. A malicious value could allow an attacker to execute arbitrary commands and potentially exfiltrate sensitive environment variables or the CLOUDFLARE_API_TOKEN itself.
[DYNAMIC_EXECUTION]: The skill implements a verification mechanism that dynamically generates a GitHub Actions workflow file at runtime and pushes it to a temporary branch in the user's repository for execution. This pattern of code generation and execution in a sensitive CI/CD environment is a high-risk practice that increases the attack surface for the user's repository.
[CREDENTIALS_UNSAFE]: The permission 'probe' mechanism involves creating a new, temporary execution context that is explicitly granted access to the production CLOUDFLARE_API_TOKEN secret. When combined with the identified command injection vulnerability in the generator script, this creates a high risk of credential theft or unauthorized access to the Cloudflare account.

deploying-cloudflare-web-application