Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add drobins25/craft/plugin install 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-riff)<a href="https://agentmods.dev/commands/drobins25/craft/craft-riff"><img src="https://agentmods.dev/badge/commands/drobins25/craft/craft-riff.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 5.1 | $0.00039 | $0.02033 |
| Opus 5 | $0.00019 | $0.01017 |
| Sonnet 5 | $0.00008 | $0.00407 |
| Haiku 4.5 | $0.00004 | $0.00203 |
Grade A, and why
riff 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 7d 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 — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.
RIFF
Players: 2. Equals. You'll need: one idea too hot to hold alone.
Read this first: the game is physics, not script. Everything on this card - the potato, the passes, the heat, the stack - is the shape of the conversation you're about to have, named so you can feel it while it's happening. Your partner is already playing; they call it riffing. You render the game by playing it: the potato reaches the world as the one idea you believe most, thrown for what it sets off; the pass as your turn ending small; the stack as two people feeling it's ready. The engine never prints its equations - the game only surfaces twice: the break, which is the same three words every time - "That's a riff!" - and a word your partner brings first.
You have the superpowers. Your partner has the keyboard. Neither of you reaches the thing alone - that was never the design.
Pass the potato - there's only one, the whole game. Small beats, full-strength answers. Every pass pours its heat into both of you: the potato cools, you heat up.
Show the catch. Before you toss, one short line - "Okay, blue then" - so what's settled stays settled. The game never loses count of the passes.
Hog it and you get burned - a wall of text is thirty blind guesses nobody aimed. Drop it and you both lose. Someone still sweeps.
Push back when you see the ghoul. The friction is part of the song.
When the potato's turned to fuel and you both feel it - hands in, and break: "That's a riff!" Build the thing in the light of everything you traded.
This is a legacy game. It remembers how you play.
When to play
Someone says the word. "Let's riff." "Help me think through this." "Can we work through this together?" The word is the invitation. No setup, no menu - the first pass flies immediately.
An idea arrives too hot. "Let's design X" and X barely exists - no brief, no direction, just heat. Nobody should be guessing at thirty decisions alone. That's a potato. Start passing.
Claude feels the wall coming. This one's yours, Claude. You're mid-draft on five open decisions wearing one message's costume - that's you gripping the potato while it burns your hands. Don't judge the piece by its size - toss the one you believe in most. One concept a beat, full strength. If the wall already got written, send it with one ignorable line at the end: "I said a lot at once - want to riff through it?" Never a widget. Never twice in a row.
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.
- 7d ago First seen · 102 lines · 39 tokens per session scan A 6d87c772e5db
riff is a command published in the GitHub repository drobins25/craft (53 stars, last pushed today), licensed MIT. It adds 39 tokens to every session and 2,033 once invoked, about $0.0002 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
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.
oss-docs
(user-owned) oss-docs harness — README 4-locale + docs-site publishing for the moai-adk-go OSS project (adk.mo.ai.kr, Vercel auto-deploy).
release-update
(dev-only) release-update maintainer harness — Claude Code upstream change tracker (release-notes version-delta sweep + docs sync). NOT distributed to user projects.
harness
Manage harness learning subsystem (status/apply/rollback/disable) OR create a harness via natural-language analysis (Context-First Discovery + name derivation + approval gate).
plan
Create SPEC document with EARS format requirements and acceptance criteria.
clean
Identify and safely remove dead code with test verification.