The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the frontcli tool from the author's official GitHub repository (github.com/dedene/frontapp-cli) via Homebrew or Go. These are verified vendor resources for the primary tool.
[COMMAND_EXECUTION]: The skill uses the frontcli command-line utility to perform its operations. This behavior is expected and matches the skill's stated purpose of managing Front platform data.
[PROMPT_INJECTION]: An indirect prompt injection surface is present. The agent reads untrusted external data (customer conversation content) via commands like frontcli conv get and frontcli conv messages. The skill does not define specific boundary markers or sanitization procedures for this ingested content, which could potentially contain malicious instructions targeting the agent's logic.
[DATA_EXFILTRATION]: While the skill facilitates data access, it is scoped to the Front platform. The instructions include security best practices, such as forbidding the inspection of keychain contents and warning against logging OAuth tokens or client secrets.

frontcli