The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and execute instructions retrieved from external task and mailbox files, creating a surface for potential instruction injection if these files are compromised or contain untrusted content.
Ingestion points: The agent reads task assignments from inbox.md and detailed task data from task-<id>.json files located in the <team_state_root>.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the content of these files as data rather than instructions.
Capability inventory: The skill possesses the ability to execute shell commands via the omx CLI tool and perform filesystem writes to state and status files.
Sanitization: The instructions do not specify any validation or sanitization logic for the content parsed from the inbox or task JSON files.

worker