transactional-email

Warn

Audited by Gen Agent Trust Hub on Oct 10, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The script scripts/send-plan-change-email.ts includes logic to access the local filesystem at ~/Library/Preferences/.wrangler/config/default.toml to extract an oauth_token. This token is used to authenticate the Cloudflare SDK for sending one-off emails.
  • [DYNAMIC_EXECUTION]: The skill provides a workflow for creating a temporary Cloudflare Worker at runtime within a tmp/ directory, involving the generation of both wrangler.jsonc and worker.ts files which are then executed using development tools.
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using wrangler and tuistory to manage local test environments and send emails via a proxied worker session.
  • [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data, such as repository names and branch strings, by interpolating them into HTML templates for email delivery.
  • Ingestion points: Data is ingested via the data parameter in the buildWelcomeEmailHtml function within SKILL.md.
  • Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the processed data.
  • Capability inventory: The skill utilizes network-capable tools including the Cloudflare send_email binding and the cloudflare SDK to transmit generated content to external recipients.
  • Sanitization: A custom escapeHtml function is implemented and required for all interpolated variables to mitigate basic HTML injection risks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 10, 2026, 10:52 AM