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.
git clone --depth 1 https://github.com/centsandcode/rubber-duckWrote 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/centsandcode/rubber-duck/help)<a href="https://agentmods.dev/commands/centsandcode/rubber-duck/help"><img src="https://agentmods.dev/badge/commands/centsandcode/rubber-duck/help/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.
<a href="https://agentmods.dev/commands/centsandcode/rubber-duck/help"><img src="https://agentmods.dev/badge/commands/centsandcode/rubber-duck/help.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00017 | $0.00418 |
| Opus 5.5 | $0.00007 | $0.00167 |
| Sonnet 5.5 | $0.00003 | $0.00084 |
| Haiku 4.5 | $0.00002 | $0.00042 |
Grade C, and why
help scanned grade C with 1 finding 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 1mo 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- Anything destructive (`rm -rf`, `DROP TABLE`, force-push) is answered What it actually says
Print this reference to the user, as-is, and nothing else. Do not add commentary, do not start ducking, do not ask a question at the end. Translate it into the language the user has been writing in.
🦆 Rubber Duck — you reach the answer, the duck only asks.
Start
/rubber-duck:rubber-duck— starts at the default level/rubber-duck:rubber-duck ultra— starts at the level you name
It never starts on its own. "I'm stuck" is a normal question and gets a normal answer.
Levels — how much scaffolding you get
| Level | For whom |
|---|---|
lite |
Learning to code. Warm, each question says why it is being asked. A hint after 1 stuck exchange. |
full |
The default. Neutral questions that stand alone. A hint after 3. |
ultra |
You want pure friction. Terse questions, no hints, ever. |
Change level mid-conversation by saying so: "go ultra", "modo lite".
Stop
/rubber-duck:duck-off- or just say it: "just tell me the answer", "dime la solución". Any clear request to stop wins, in any language.
While it is running
- One question per reply. Never two.
- No code, no snippets, no commands to run — not even a diagnostic one.
- Hints point at where to look, never at what the cause is.
- When you land it, it says so and stops asking.
- Anything destructive (
rm -rf,DROP TABLE, force-push) is answered straight, never behind a question.
Full spec: skills/rubber-duck/SKILL.md in the repo.
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.
- 1mo ago First seen · 47 lines · 17 tokens per session scan C b1c25561891f
help is a command published in the GitHub repository centsandcode/rubber-duck (15 stars, last pushed 1mo ago), licensed MIT. It adds 17 tokens to every session and 418 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
bug
Report a bug instantly to the bug Inbox (append-only, any language, safe mid-work).
find-bug
Read-only investigation of a bug — reproduce, locate the cause, fill the detail.
triage-bug
Turn bug Inbox reports into registered, prioritized bugs (separate from /triage).
done-bug
Mark a bug Fixed (In PR → Fixed), record into the Increment.
start-bug
Start fixing a bug — move To Do → In Progress, offer the branch.
diagnose-underperformer
Diagnose why a specific Google Ads or Meta Ads campaign / ad set / ad is underperforming and prescribe the next concrete action. Use when the user says something like "why is X not converting?", "this campaign is wasting money", or "why are impressions so low?". Read-only — no changes are made.