The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the majal CLI tool to perform various data processing tasks, including scan, stats, fix, and sample on user-provided local files.
[PROMPT_INJECTION]: The skill processes Arabic training datasets (data.jsonl) which are untrusted external inputs. This presents an indirect prompt injection surface where malicious instructions embedded in the data could attempt to influence the agent's behavior during analysis or repair.
Ingestion points: The skill ingests data from .jsonl files as seen in the command examples in SKILL.md.
Boundary markers: There are no explicit boundary markers or system instructions provided to ensure the agent ignores embedded instructions within the processed data.
Capability inventory: The skill executes shell commands (majal) on the target files, allowing it to read and modify the content.
Sanitization: The instructions do not specify any sanitization, filtering, or validation steps for the content of the training data before it is processed.

majal