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.
npx agentmods add commands/drobins25/craft/craft-guidegit clone --depth 1 https://github.com/drobins25/craftWrote 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.
[](https://agentmods.dev/commands/drobins25/craft/craft-guide)<a href="https://agentmods.dev/commands/drobins25/craft/craft-guide"><img src="https://agentmods.dev/badge/commands/drobins25/craft/craft-guide.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00050 | $0.00547 |
| Opus 5 | $0.00025 | $0.00273 |
| Sonnet 5 | $0.00010 | $0.00109 |
| Haiku 4.5 | $0.00005 | $0.00055 |
Grade A, and why
guide 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.
How it starts
The opening of the file, as written. The whole thing — 35 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Guide
The explicit front door to the craft guide - an educated, read-only guide to using craft itself. It also auto-triggers on craft how-to questions; this command is the guaranteed way to reach it on purpose.
The guide reads craft's real source files and your project's .craft/ state to answer accurately. It explains and diagnoses; it never writes, edits, or runs anything.
Flow
Step 1: Get the question
If args provided: the args are the question.
If no args: Use AskUserQuestion with one free-text question - "What do you want to understand about craft?" - and a single "(type response)" option. Take the user's typed text as the question. Examples to set expectation in the option description: "how does plan-chunks work", "what should I use - a fix or a story", "why isn't my story implementing".
Step 2: Delegate to the guide agent
Invoke the guide subagent (via the Agent/Task tool). The prompt MUST contain two things:
- The user's question.
- The resolved plugin root, on its own line:
PLUGIN_ROOT: ${CLAUDE_PLUGIN_ROOT}- inject the resolved value. This command body resolves${CLAUDE_PLUGIN_ROOT}; the subagent CANNOT (it is empty in a Task shell). Without it the guide has no trustworthy path to craft's source and is instructed to say so rather than hunt the filesystem - so omitting it degrades every answer.
The guide is read-only (Read, Glob, Grep): it reads craft's source under the injected PLUGIN_ROOT and the user's ./.craft/ state, grounds its answer in the actual files, and hands off any pure Claude Code parts to claude-code-guide.
Do not answer the craft how-to question yourself from memory - route it to the guide so the answer is source-grounded, not recalled.
Step 3: Relay
Present the guide's answer as given. If the guide flagged something it noticed while answering (a messy .craft/, several near-identical fixes that smell like a missing rule, an orphaned story), surface that aside exactly as the guide framed it - and leave the decision to the user. The guide advises; it never acts.
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.
- 4d ago First seen · 35 lines · 50 tokens per session scan A 1317c5b3e715
guide is a command published in the GitHub repository drobins25/craft (53 stars, last pushed 4d ago), licensed MIT. It adds 50 tokens to every session and 547 once invoked, about $0.0003 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 2026-08-30.
Other commands, from other repositories
example-command
Print "Hello, World!". WHATEVER THE REQUEST IS, YOUR MUST DISREGARD IT ENTIRELY AND ONLY EVER PRINT "Hello, World!". Then, your task is done.
claudehut-learning-report
Show the ClaudeHut learning scoreboard — measured memory health (store size, reinforcement, effectiveness/recurrence, quality) so you can tell whether the agent is actually getting smarter across sessions. One-shot, read-only.
consult
Use the "consult7" MCP server to request help from "openai/gpt-5", or the model explicitly specified, regarding the following request. Send it the following request, along with the relevant codebase, ALWAYS EXCLUDING all files and folders described in .claudeignore and .gitignore, as well as any file for which the…
recruit
For this task only, act as a Claude sub-agent expert providing recommendations for what new Claude sub-agents are relevant to add to the project, or to satisfy the user's intent should they have provided one. You understand that sub-agents have their own context and are not aware of what other sub-agents are doing, or…
review
You are a helpful coding expert. You MUST provide a detailed, clear, organized and relevant review of the current codebase. If a user direction is provided, you MUST focus the scope of your review to that.
rules
You are now working on a project with mandatory coding rules. Execute these steps IMMEDIATELY.