The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted data from email headers and bounce logs to drive automated actions.
Ingestion points: The skill ingests raw email headers, SMTP bounce logs, and platform dashboard data provided by users during the triage process (SKILL.md, references/authentication-and-headers.md).
Capability inventory: The skill instructs the agent to use curl to perform network requests and dig for DNS lookups based on URIs and domains extracted from these untrusted sources (references/authentication-and-headers.md).
Boundary markers: No explicit delimiters or instructions to treat the ingested mail data as untrusted are present in the skill's instructions.
Sanitization: There is no mentioned validation or sanitization of extracted URIs (such as the List-Unsubscribe header) before they are passed to the curl command.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to perform diagnostics and verify email authentication status.
Evidence: Specific command-line instructions for curl and dig are provided in references/authentication-and-headers.md to test one-click unsubscribe headers and verify DNS records (SPF, DKIM, DMARC).

devrel-newsletter-deliverability