The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data when reading or listing emails, which presents a surface for indirect prompt injection where instructions within an email could attempt to manipulate the agent.
Ingestion points: himalaya list and himalaya read in SKILL.md
Boundary markers: Absent
Capability inventory: himalaya send, himalaya reply, and shell command execution
Sanitization: Not explicitly defined
[COMMAND_EXECUTION]: The skill's primary operations are performed through the execution of shell commands, including tool installation and email management.
[CREDENTIALS_UNSAFE]: Documentation provides examples for configuring email accounts, which involves placing sensitive app-passwords in configuration files or passing them as command-line arguments to curl.
[EXTERNAL_DOWNLOADS]: The skill references the installation of the himalaya CLI client through trusted and well-known sources including Homebrew, Cargo, and GitHub releases.

email-tools