checklist

name	checklist
description	Run a **repeatable checklist / to-do list** from a markdown file of `- [ ]` items — execute the open items, tick them `[x]`, and write the file back. Re-running resumes only the still-open items, so the file IS the durable, repeatable state. USE for a generic user to-do that should be runnable again and again: "run my TODO.md", "esegui la checklist", "go through CHECKLIST.md", "do the open items in release-steps.md". Each `- [ ]` line is itself the instruction for that item (optionally refined by --prompt). This is thin sugar over /agentflow:foreach with `--checkbox`: same engine, same durability and cross-turn resume, friendlier defaults. For "apply one operation to a JSON list / folder / glob" use /agentflow:foreach directly; to first GENERATE the list use /agentflow:enumerate; for one combined output use /agentflow:reduce.
allowed-tools	Bash, Read, Write, Edit, Glob, Grep, Agent
argument-hint	<checklist.md> [--prompt "<how to do each item>"] [--serial] [--model haiku|sonnet|opus] [--execution main-thread|subagent] [--dry-run]

/agentflow:checklist

Make it visible: in one line say you're starting a checklist run (file + run-id). /agentflow:board then lists it on disk — the audit trail. All paths are relative to the workspace root.

You are running a repeatable checklist. This is /agentflow:foreach --checkbox with defaults — follow the foreach skill's orchestration (${CLAUDE_PLUGIN_ROOT}/skills/foreach/SKILL.md); this file only fixes the sugar.

Resolve arguments

• File (required): the first positional, else the user's named file, else TODO.md in the workspace root if present. It must contain markdown task lines (- [ ] ..., completed ones - [x] ...).
• Per-item operation (--prompt): default is "Complete this checklist task. The task text is in the item's data. Do it end-to-end using the available tools, then report concisely what you did and any follow-up. If the task is ambiguous or cannot be completed, say so and leave it for review." Override with --prompt (e.g. --prompt "Draft the section described by each item").
• --serial → run items in list order, one at a time (sequential to-do). Default is parallel chunks.
• --model / --execution / --dry-run pass through to foreach (--dry-run = --validate-only + report the open-item count without executing).

Run it (defaults wired)

1. Init a run (id like checklist-<file-stem>-<date>):

   node "${CLAUDE_PLUGIN_ROOT}/dist/state/foreach.js" init <run-id> --checkbox <file> --prompt "<operation>" [--serial]
   

   --checkbox pre-marks existing [x] lines as done, so only open items are scheduled (this is what makes re-runs resume cleanly). For --dry-run, add --validate-only and stop here, reporting total/pending.
2. Drive the run exactly as the foreach skill describes: claim a chunk → dispatch each item to a subagent (or do it inline for main-thread) with the operation + the item's task text → record results with complete/complete-batch. The state file is the only source of truth; subagents never write state.
3. Write back: when items reach a terminal state, reflect them onto the file:

   node "${CLAUDE_PLUGIN_ROOT}/dist/state/foreach.js" view <run-id> --checkbox <file>
   

   Done items become - [x]; failed/blocked ones stay - [ ] and are reported for the next run.
4. Close: print a one-line summary (done / failed / remaining) and the run-id. Re-running /agentflow:checklist <file> later picks up exactly the still-open items.

Notes

• Repeatable by design: the checkbox file is the state. Nothing else is needed to resume.
• Cross-turn durability and the Stop-hook auto-resume come for free from the foreach engine.
• To stop a long run, create the foreach --stop-file; to cap cost, pass --max-usd (see foreach).