The Agent Skills Directory

[SAFE]: The skill is a legitimate utility for organizing multi-agent tasks. It does not contain any code or instructions that violate security policies.
[COMMAND_EXECUTION]: The script scripts/task_board.py facilitates the creation of project plans. It uses standard Python libraries to parse command-line arguments and output formatted JSON or Markdown. The script does not execute arbitrary commands, access the network, or interact with sensitive files.
[PROMPT_INJECTION]: The prompt template in references/prompt-template.md uses structured instructions to define the 'Coordinator' role. It does not include directives to bypass safety filters, ignore system instructions, or engage in malicious role-play.
[INDIRECT_PROMPT_INJECTION]: While the skill processes external input (task goals and worker names) and interpolates them into a prompt template, this is the intended functionality of a coordinator skill. The skill lacks dangerous tools (such as network or high-privilege file access) that would make this interpolation risky. Boundary markers are clearly defined in the template structure.

swarm-coordinator