help

help is a command for coding agents from JonyanDunh/claude-code-watchdog. It costs 7 tokens per session (2,379 once invoked), scanned A, original, Apache-2.0.

A help command for Watchdog, a Claude Code loop that repeats a prompt while each turn is still changing project files.

In plain words
What is it for?
Use it to learn Watchdog's commands, repeated work cycle, file-change check, and stopping rules.
Why use it?
It explains how Watchdog decides whether work is continuing and when the loop stops, including its iteration limit and manual stop command.

Command

Part of the watchdog plugin — 3 commands, 1 hook shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/jonyandunh/claude-code-watchdog/help
Clone the repo
git clone --depth 1 https://github.com/JonyanDunh/claude-code-watchdog

Or install watchdog, the plugin that ships this one along with the rest of its 3 commands, 1 hook.

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.

agentmods badge for help

README.md
[![agentmods](https://agentmods.dev/badge/commands/jonyandunh/claude-code-watchdog/help.svg)](https://agentmods.dev/commands/jonyandunh/claude-code-watchdog/help)
Your own site
<a href="https://agentmods.dev/commands/jonyandunh/claude-code-watchdog/help"><img src="https://agentmods.dev/badge/commands/jonyandunh/claude-code-watchdog/help.svg" alt="Measured on agentmods" height="20"></a>
Per session 7 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,379 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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.

ModelPer sessionOnce invoked
Fable 5 $0.00007 $0.02379
Opus 5 $0.00003 $0.01189
Sonnet 5 $0.00001 $0.00476
Haiku 4.5 $0.00001 $0.00238

Measured 4d ago against content hash 3d8ffcc4ba8a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

help 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 4d 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.

commands/help.md · 119 lines

How it starts

The opening of the file, as written. The whole thing — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Watchdog Plugin Help

Please explain the following to the user in clear, friendly language. Do not dump the markdown verbatim — synthesize and present it.

What is Watchdog?

Watchdog is a self-referential loop for Claude Code. You give it a prompt once, and it re-feeds that same prompt to Claude after every turn until the task genuinely stops producing file edits. The agent is not told a loop is running, so it cannot fake a completion signal to escape early.

Core mechanic:

  1. You run /watchdog:start "<prompt>" [--max-iterations N]
  2. Claude works on the task, modifying files
  3. Claude's turn ends — the Stop hook intercepts
  4. A headless Claude Haiku classifier inspects every tool invocation from that turn and decides: did any of them directly modify a project file?
  5. If yes → re-feed the original prompt as a new user turn, loop continues
  6. If no (or --max-iterations reached, or /watchdog:stop invoked) → exit cleanly

The "self-reference" means each iteration sees the previous iteration's file edits and git state, so Claude builds on its own prior work without any external orchestrator.

Why the Haiku classifier?

Instead of a hard-coded tool-name whitelist (which misses Bash(sed -i …), MCP SQL writes, and other indirect mutations), Watchdog asks a headless Haiku instance to read every tool invocation's full input and judge semantically whether a project file was modified. The verdict is a single token: FILE_CHANGES or NO_FILE_CHANGES. Side effects that don't touch project files — running containers, remote DB writes, network calls — are correctly ignored.

Available commands

/watchdog:start "<PROMPT>" [--max-iterations N] [--exit-confirmations N] [--prompt-file <path>] [--watch-prompt-file] [--no-classifier]

Start a Watchdog in the current session.

Usage:

/watchdog:start "Refactor services/cache.ts to use the new API. Iterate until pnpm test:cache passes." --max-iterations 20
/watchdog:start "Add tests for auth.ts until coverage hits 80%."
/watchdog:start --prompt-file ./tmp/my-big-prompt.txt --max-iterations 20
/watchdog:start "Refactor cache.ts" --exit-confirmations 3 --max-iterations 20
/watchdog:start --prompt-file ./tmp/task.md --watch-prompt-file --max-iterations 20
/watchdog:start "Build until I /watchdog:stop" --no-classifier

Read the full file on GitHub · 119 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.

  1. 4d ago First seen · 119 lines · 7 tokens per session scan A 3d8ffcc4ba8a

Subscribe to this mod's changes

help is a command published in the GitHub repository JonyanDunh/claude-code-watchdog (4 stars, last pushed 4mo ago), licensed Apache-2.0. It adds 7 tokens to every session and 2,379 once invoked, about $0.0000 per session on Opus 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 2026-08-31.