The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses instructional language to manage research modes and dependencies. No attempts to override safety filters or extract system instructions were identified. While the skill reads local files created by other processes, its capability is limited to routing and status reporting, minimizing risks of indirect injection impact.\n- [DATA_EXFILTRATION]: The skill reads project-specific files in 'research-memory/' and 'brand-memory/' to determine research completeness and staleness. This data access is localized and necessary for orchestration; no patterns for transmitting this data to external domains or endpoints were found.\n- [COMMAND_EXECUTION]: The skill serves as a controller for other tools but does not directly execute shell commands, spawn subprocesses, or invoke system-level APIs.\n- [EXTERNAL_DOWNLOADS]: No network requests, external resource fetching, or downloads are initiated by this skill's logic.

research-orchestrator