The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool for routine file system operations, including the creation of a task directory using mkdir -p tasks and determining file counts via ls tasks/{type}-* | wc -l. These commands are localized to the working directory.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted data to populate task templates.
Ingestion points: Untrusted data enters the context through user-supplied feature descriptions via the AskUserQuestion tool and task lists from the TodoWrite integration as described in SKILL.md.
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions to prevent the agent from potentially following commands embedded within the input descriptions.
Capability inventory: The skill utilizes Bash, Write, and Read tools, which define the potential impact if an injection were to occur during the processing phase.
Sanitization: No input validation or sanitization of the feature descriptions or task data is specified before the agent processes them into structured files.

create-task-files