A command that keeps a background watch running until another seat's turn arrives, then wakes the coding agent. A seat is a declared participant in the project's turn-taking workflow.

Arming a role and workflow definition for background turn detection, optionally scoped to a particular workflow run.

It avoids a foreground wait that can time out and reduces the need for an operator to restart the agent manually when work is ready.

How JanoshMoshiri/MarkdownLLM configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to JanoshMoshiri/MarkdownLLM. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/JanoshMoshiri/MarkdownLLM/main/.claude/commands/watch.md

Clone the repo

git clone --depth 1 https://github.com/JanoshMoshiri/MarkdownLLM

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/commands/janoshmoshiri/markdownllm/watch/github.svg)](https://agentmods.dev/commands/janoshmoshiri/markdownllm/watch)

Your own site

<a href="https://agentmods.dev/commands/janoshmoshiri/markdownllm/watch"><img src="https://agentmods.dev/badge/commands/janoshmoshiri/markdownllm/watch/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/commands/janoshmoshiri/markdownllm/watch"><img src="https://agentmods.dev/badge/commands/janoshmoshiri/markdownllm/watch.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 51 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 823 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00051	$0.00823
Opus 5.5	$0.00020	$0.00329
Sonnet 5.5	$0.00010	$0.00165
Haiku 4.5	$0.00005	$0.00082

Measured 3d ago against content hash 654979207a9b, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

watch scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/commands/watch.md · 61 lines

How it starts

The opening of the file, as written. The whole thing — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Why this is a command. The standing watch (mdllm watch) detects the other seat's turn and exits 0 on yours; what rises on that exit is the harness's act, and until 2026-10-30 Claude Code had no recorded binding for it. The loop definitions said "hold the blocking watch inside your turn", which in Claude Code dies at the tool timeout, so the record shows turns picked up "through the operator's chat prompt, not an automatic wake". This command is the binding: the watch runs as a background command, which Claude Code keeps running across turns and re-invokes you on when it exits. Never run it in the foreground.

Arguments: $ARGUMENTS — <role> <definition> [run], this seat's actor name, the loop definition's id, and optionally the workflow-run the conversation is scoped to.

Arm

Run the watch in the background (the Bash tool's run_in_background), from the workspace root, through the launch route the entry file names:

mdllm watch . --definition <definition> [--role <role>] [--run <run>] --exit-on-wake

Before arming, confirm the seat is declared (mdllm seat .): the watch rings a declared seat only for its own unpublished commits, and tells an undeclared one it cannot narrow whose turn never left. Then tell the operator in one line what is armed and against which ref, and end your turn. The watch's startup line (watching … as [role] … silence means nothing moved) is the receipt; already watching … (pid N) with exit 3 means another instance holds this role in this clone — do nothing.

Rise

When the background command exits, the harness re-invokes you with its output. Read the exit code before anything else:

Exit	Meaning	What you do
`0`	Your turn: a thing moved to a stage this role acts at, a note landed on one you hold (`a note landed on …` — reread it, the turn did not move), or your own turn never left	Take the turn as the loop definition says — read the thing at the remote head, do the work, move the token, commit, publish — then re-arm with the same command, in the background, before ending the turn
`3`	Another instance of this role is already watching	Nothing. The other instance is the watch
`1`	A poll failed in `--once` mode	Report the line; re-arm
`2`	Refused to arm	Fix the declaration it names; do not retry blindly

Read the full file on GitHub · 61 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

3d ago First seen · 61 lines · 51 tokens per session scan A 654979207a9b

Subscribe to this mod's changes

watch is a command published in the GitHub repository JanoshMoshiri/MarkdownLLM (5 stars, last pushed 4d ago), licensed MIT. It adds 51 tokens to every session and 823 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-03.

Building agents Productivity

watch next to 6 commands from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
watchJanoshMoshiri/MarkdownLLM	Command	5	4d ago	MIT	A	51	Claude Code
wsdanielmiessler/LifeOS	Command	20k	yesterday	MIT	A	0	Claude Code
cancel-ralphanthropics/claude-plugins-official	Command	46k	yesterday	Apache-2.0	A	4	Claude Code
engage.actionshypnguyen1209/offensive-claude	Command	412	15d ago	MIT	A	11	Claude Code
focusData-Wise/claude-plugins	Command	7	yesterday	MIT	A	0	Claude Code
stuckData-Wise/claude-plugins	Command	7	yesterday	MIT	A	0	Claude Code
quickstartorq-ai/assistant-plugins	Command	6	2d ago	MIT	A	23	Claude Code

MarkdownLLM: Command for Claude Code