The Agent Skills Directory

[COMMAND_EXECUTION]: The skill details the use of platform-specific commands for subagent management (Claude Code Task/Agent and Codex) and dynamic workflows. It includes a standard installation command npx skills@latest add tylergibbs1/dispatch associated with the author's identity.\n- [EXTERNAL_DOWNLOADS]: The documentation references Context7 via GitHub and an MCP server URL (https://mcp.context7.com/mcp) to ensure subagents use up-to-date library documentation. This is a functional requirement for verifying technical versions.\n- [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection. Untrusted data enters the context during repository scouting (SKILL.md) and subagent reporting (SKILL.md). The skill mitigates this by mandating self-contained briefs, using checklists for state management, and requiring manual verification of all subagent outputs before integration. Subagents possess the capability to perform workspace writes and tool execution.\n- [SAFE]: No malicious activity was detected. The orchestration process includes clear boundaries for subagents, mandatory result verification, and explicit user consent for scaling to large workflows via the ultracode flag. No obfuscation, persistence mechanisms, or unauthorized credential access attempts were found.

orchestrating-subagents