The Agent Skills Directory

[PROMPT_INJECTION]: The skill explicitly directs the agent to "Operate fully autonomously" and "Never ask the user to choose or pause for confirmation," bypassing standard human-in-the-loop safety protocols.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting potentially untrusted data to generate task prompts.
Ingestion points: Data is read from .noodle/mise.json (backlog items and event payloads) and files within brain/plans/.
Boundary markers: Absent; there are no instructions to the model to treat external data as untrusted or to ignore embedded commands.
Capability inventory: The skill can perform filesystem writes (.noodle/orders-next.json) and execute CLI commands via noodle.
Sanitization: Absent; external titles, descriptions, and event data are interpolated directly into task prompts and the extra_prompt field.
[COMMAND_EXECUTION]: The skill utilizes the noodle CLI to perform operations such as noodle schema mise and noodle adapter run backlog add.

schedule