The Agent Skills Directory

[SAFE]: The skill serves as a technical manual for orchestrating complex fan-out tasks. It documents patterns for task decomposition, manifest generation, and result synthesis without introducing malicious payloads.
[DYNAMIC_EXECUTION]: The workflow describes using the execute_code tool to run Python scripts for deterministic orchestration tasks, such as generating manifest files and managing loops. This is used as a performance and context-management optimization rather than for arbitrary code execution.
[COMMAND_EXECUTION]: The instructions involve the use of the terminal tool by subagents for processing individual task units. These operations are described within the context of standard file system and repository management.
[DATA_EXFILTRATION]: The skill uses local /tmp directories with unique run identifiers to store intermediate results and manifests. No patterns for exfiltrating sensitive data to external domains were identified.

dynamic-workflow