The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is explicitly instructed to execute shell commands to terminate training processes. This includes running a user-provided stop_command from the context or autonomously determining commands to stop tmux sessions, local/remote processes, or scheduler jobs using Bash(*).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from training logs and WandB metrics to make decisions, which creates a potential injection surface.
Ingestion points: External log files (via fallback log path) and WandB run data as described in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions for the log content being analyzed.
Capability inventory: The skill possesses Bash(*), Read, Write, Edit, and Grep capabilities, intended for metric extraction and process termination.
Sanitization: There is no evidence of sanitization or filtering of the ingested log content before it is processed by the agent's logic.

training-check