The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution by interpolating the user-provided $ARGUMENTS into a bash command string: parallel-cli research status "$ARGUMENTS" --json. This allows for potential command injection if a user provides input containing shell metacharacters like ;, &&, or |. While the allowed-tools field restricts commands to the parallel-cli prefix, this check can often be bypassed if the malicious input is appended after the initial command components.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input which could contain malicious instructions.
Ingestion points: The $ARGUMENTS variable in SKILL.md receives input directly from user requests.
Boundary markers: No delimiters or safety instructions are used to separate the user-provided data from the command structure.
Capability inventory: The skill utilizes the Bash tool to execute commands on the system.
Sanitization: No input validation, filtering, or escaping is performed on the $ARGUMENTS variable before it is passed to the shell.

status