The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/mailhog_manager.sh script employs sudo to move the downloaded MailHog binary into /usr/local/bin. This is a privilege escalation pattern where the skill seeks root-level write permissions on the host system.
[COMMAND_EXECUTION]: The scripts/mailhog_manager.sh script modifies the user's ~/.bashrc file to update the PATH environment variable. This is a persistence mechanism that makes permanent changes to the user's shell environment.
[EXTERNAL_DOWNLOADS]: The script scripts/mailhog_manager.sh downloads the MailHog binary from GitHub's release page. While GitHub is a well-known service and the repository is legitimate, the automated downloading and positioning of binary executables is a high-risk activity for an AI agent skill.
[PROMPT_INJECTION]: The skill provides functionality to read, search, and process captured emails, which serves as a significant surface for indirect prompt injection. Malicious instructions embedded in emails could influence the agent's subsequent actions.
Ingestion points: Email data is ingested via the MailHog API (/api/v1/messages) as referenced in references/api-endpoints.md and scripts/test_email_workflow.sh.
Boundary markers: The skill does not implement boundary markers or instructions to isolate untrusted email content from the agent's primary instructions.
Capability inventory: The skill includes several scripts capable of network communication, file system access, and shell command execution, providing multiple avenues for exploitation if an injection succeeds.
Sanitization: There is no evidence of sanitization or content validation performed on the email data before it is presented to the agent.

mailhog