Use when Claude or Codex subagents should do part of the work from timon - one run with a validated answer, fan-out over many items, author and reviewer loops, isolated git worktrees, budgets, and reading past runs. Runs spend tokens. Load the timon skill…
Use when earlier Claude Code or Codex work may already hold the answer - find past sessions, search every session's text, read a transcript around a match, summarize a session, or tell which session you are in. Load the timon skill first for the basics.
Use timon when a task is easier as a few lines of code than as many tool calls - reading and reshaping files and JSON, querying with JMESPath, running programs and concurrent work, or scripting in Lua or Fennel - and to run a project's tasks (timon tasks,…
Use timon when a task is easier as a few lines of code than as many tool calls - reading and reshaping files and JSON, querying with JMESPath, running programs and concurrent work, or scripting in Lua or Fennel - and to run a project's tasks (timon tasks,…
Use when defining or running timon tasks - a project's or your own tasks.lua, timon tasks, timon NAME, typed options and --help, dependencies and --jobs, task.run, --dry-run, words after --, aliases, groups, confirm and platforms, and tasks as Claude Code and…
Use when a task needs more than one agent run or a check on an agent's work - verify findings, vote, pick the best of several, revise until a gate passes, retry, or run steps in dependency order with timon's flow patterns. Load timon-agents first for…
Use when writing timon scripts to keep - named scripts in ~/.config/timon or a project's .timon, shared lib modules, settings in init.lua, structured concurrency with scope, and Fennel. Load the timon skill first for the basics.