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 skills add harnessprotocol/harness-kit --skill rubber-duckygit clone --depth 1 https://github.com/harnessprotocol/harness-kitWrote 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/skills/harnessprotocol/harness-kit/rubber-ducky)<a href="https://agentmods.dev/skills/harnessprotocol/harness-kit/rubber-ducky"><img src="https://agentmods.dev/badge/skills/harnessprotocol/harness-kit/rubber-ducky/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/skills/harnessprotocol/harness-kit/rubber-ducky"><img src="https://agentmods.dev/badge/skills/harnessprotocol/harness-kit/rubber-ducky.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00186 | $0.02069 |
| Opus 5.5 | $0.00074 | $0.00828 |
| Sonnet 5.5 | $0.00037 | $0.00414 |
| Haiku 4.5 | $0.00019 | $0.00207 |
Grade A, and why
rubber-ducky 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 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.
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 — 193 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Rubber Ducky
A built-in critic. Before committing to a non-trivial change, you stop, articulate your current thinking, and have it scrutinized by independent reviewers running on different models than the one driving this session. Unlike a real rubber duck, this one talks back: it returns a structured critique you can act on.
Why this works
Rubber-duck debugging works because articulating your reasoning forces the gaps into view. This version adds two things a desk toy can't:
- It talks back. The reviewers return concrete, categorized feedback — not silence.
- It doesn't share your blind spots. The reviewers run on the two Claude models you are not using right now, with a clean context. A model that didn't produce the work is far more likely to see what's wrong with it. You get genuinely independent perspectives, not a re-run of the same reasoning that produced the plan.
The single highest-leverage moment is after you've planned a change but before you've written the code. A design flaw caught here costs a paragraph to fix; caught after implementation it costs a rewrite.
rubber-ducky vs /review
Both are read-only critics, but they fire at different stages:
- rubber-ducky pressure-tests your own in-progress thinking — a plan, a design, an approach, tests you just wrote — usually before or during implementation, using cross-model critics.
- /review evaluates a finished change — a branch diff or a PR — after the code exists.
Reach for the duck while the decision is still cheap to change; reach for /review when the work is done.
When to consult the duck
Consult it at high-leverage moments, not only when stuck:
- After planning a non-trivial change, before implementing it. Highest leverage. Corrections are cheapest here.
- Mid-implementation on complex or multi-file work, to check for blind spots before you're in too deep.
- After writing tests, to validate the coverage is real and the behavior actually satisfies the original request — not just that the tests pass.
- Reactively, when you hit repeated failures or unexpected results. Get an independent analysis instead of retrying the same approach a fourth time.
- Whenever the user explicitly asks ("rubber duck this", "/rubber-ducky", "poke holes in this").
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 193 lines · 186 tokens per session scan A d702a89035bc
rubber-ducky is a skill published in the GitHub repository harnessprotocol/harness-kit (10 stars, last pushed today), licensed Apache-2.0. It adds 186 tokens to every session and 2,069 once invoked, about $0.0007 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 2026-08-31.
Other skills, from other repositories
audit
Project health audit and health check — architecture, performance, tests, dependencies, code quality. Use when assessing overall project health, before releases, or after refactors.
skeptical-triage
Reusable 3-round self-challenge + arbiter pattern for filtering false positives from findings/verdicts. Use when the cost of a false-positive gate block exceeds the cost of 4 extra LLM turns.
anti-patterns
Catalogue of known SDLC anti-patterns that greatcto agents must actively reject when reviewing architecture, plans, code, or post-mortems. Used by architect (pre-impl), pm (planning), senior-dev (impl), l3-support (post-incident).
observability-baseline
Scaffold-time observability so a shipped product is not blind in prod from day one — error capture (Sentry), request-id structured logging, and /healthz + /readyz endpoints. stack-baseline pins Sentry but nothing wires it; this is the wiring. Loaded by app-scaffolder (bake into the scaffold), infra-provisioner (prod…
ia-verification-before-completion
Enforces fresh verification evidence before any completion claim. Use when about to claim "tests pass", "bug fixed", "done", "ready to merge", handing off work, or before editing when a request has ambiguous scope.
debugging
Systematic root-cause debugging: reproduce, investigate, hypothesize, fix with verification. Use when asked to "debug this", "fix this bug", "why is this failing", "troubleshoot", or mentions errors, stack traces, broken tests, flaky tests, regressions, or unexpected behavior.