The Agent Skills Directory

[COMMAND_EXECUTION]: Direct interpolation of user input into shell commands. In Step 2, the skill instructs the agent to execute mise x -- backlog task create "<title>" where <title> is a user-controlled string. This pattern is vulnerable to command injection if the input contains shell-active characters like ;, &, |, or $(...).
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data and uses it in privileged operations.
Ingestion points: User input provided via prompt for the task title, description, and acceptance criteria.
Boundary markers: None identified in the shell command construction or markdown template interpolation.
Capability inventory: Bash for CLI operations, Write/Edit for file system access.
Sanitization: There are no instructions for the agent to sanitize or escape user input before usage in commands or file writing.

task-create