samykabu/agent-tools
Delegate one task to an external agent CLI and get back a measured, normalised result; and keep a live, deterministic progress report for a Markdown plan or Spec Kit feature, with native Claude Code and Codex hooks
Delegate a task to an external agent CLI harness — Claude Code, OpenAI Codex, OpenCode, GitHub Copilot, or Pi — and get back a normalised result with status (successful / failed / abandoned), a summary, what actually changed on disk measured against git, and total input and output token usage. Use when asked to "use codex to ...", "delegate this to opencode", "have another agent do X", "run this task on a different model", to get a second opinion from another harness, or to compare harnesses on the same task.
Keep a live, deterministic progress report for a Markdown plan checklist or a Spec Kit feature. Collects task state from the plan file, hook events from Claude Code or Codex, and lane, blocker and decision facts you record, then renders a browser view and a Markdown export with no agent spent on upkeep. Use when asked to "track progress on this plan", "show a progress dashboard", "report which agents work on which tasks", to install progress hooks, or to record a lane assignment, blocker, decision or note.