The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes basic shell utilities such as date, ls, and mkdir for operational tasks. These commands facilitate the creation of project-specific documentation folders (docs/afk-exploration/) and track the duration of research cycles against user-defined limits. These actions are aligned with the skill's purpose and are performed transparently.
[SAFE]: The orchestration logic dispatches subagents using a structured 'brief' that includes explicit 'Scope' and 'Out of scope' constraints. This design pattern effectively bounds the subagents' actions and provides a layer of protection against indirect prompt injection from external research sources by preventing the orchestrator from directly processing subagent findings in its own context.
[DATA_EXPOSURE]: The skill reads project metadata and codebase structure to organize its research output. It adheres to a project-scoped directory tree and does not attempt to access sensitive system files or exfiltrate data to unauthorized external endpoints. No hardcoded credentials or unsafe network operations were detected.

afk-exploration