The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingest and processes task content from external todo.txt files, which may be influenced by untrusted data.
Ingestion points: Task data is read from $HOME/todo.txt and repository-specific todo.txt files using the tuxedo ls --json command.
Boundary markers: While the CLI uses JSON for output formatting, the instructions lack explicit boundary markers or warnings to prevent the agent from obeying instructions embedded within the task descriptions themselves (the raw field).
Capability inventory: The agent has the ability to execute filesystem mutations via the tuxedo CLI, including adding, deleting, and archiving entries.
Sanitization: The skill does not describe any sanitization or validation logic to filter potentially malicious instruction strings from the task data before the agent processes them for upkeep or reporting.
[COMMAND_EXECUTION]: The skill relies on executing the tuxedo CLI tool to perform task management operations. The instructions use absolute paths and environment variable overrides (TODO_FILE, DONE_FILE) to manage file access, which is consistent with the skill's primary purpose.

do-todo