The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution using the kai-brief and kai-gate utilities. User input from arguments like {format} and {keyword} is used to construct the command strings executed by the agent.
[PROMPT_INJECTION]: The {keyword} parameter is vulnerable to indirect prompt injection. A malicious user could provide a keyword containing shell metacharacters (e.g., backticks or semicolons) to escape the intended command and execute unauthorized shell instructions.
Ingestion points: The {keyword} argument used in the shell command in SKILL.md.
Boundary markers: No boundary markers or 'ignore' instructions are present for the user input.
Capability inventory: Capability to execute shell commands via kai-brief and kai-gate, and write files to the ~/.kai-marketing/ directory.
Sanitization: No sanitization, escaping, or validation of the {keyword} input is described in the skill instructions.

marketing-sprint