The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides the aiter terminal write command which allows an agent to send arbitrary strings to an active terminal. Because the skill automatically appends a carriage return, these strings are immediately executed as shell commands, granting the agent full command execution capabilities.
[DATA_EXFILTRATION]: The aiter terminal read command allows the agent to retrieve the entire text buffer of any terminal. This creates a risk of exposing sensitive information, such as passwords, API keys, or environment variables that may be visible in the terminal history.
[EXTERNAL_DOWNLOADS]: The skill depends on the external aiter CLI tool and the AiTer desktop application being pre-installed on the system. It also references the use of jq for parsing JSON output.
[PROMPT_INJECTION]: The skill has a significant surface for indirect prompt injection as it reads and acts upon the output of external processes.
Ingestion points: Terminal output via aiter terminal read in SKILL.md.
Boundary markers: Absent. The agent reads raw terminal lines with no delimiters between command output and potentially malicious data.
Capability inventory: Full terminal control (aiter terminal write), project management, and file server orchestration.
Sanitization: ANSI escape codes are stripped, but no semantic sanitization or instruction filtering is applied to the terminal output.

aiter-control