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 ARA-Labs/Agent-Native-Research-Artifact --skill research-fuzzergit clone --depth 1 https://github.com/ARA-Labs/Agent-Native-Research-ArtifactWrote 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/ara-labs/agent-native-research-artifact/research-fuzzer)<a href="https://agentmods.dev/skills/ara-labs/agent-native-research-artifact/research-fuzzer"><img src="https://agentmods.dev/badge/skills/ara-labs/agent-native-research-artifact/research-fuzzer/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/ara-labs/agent-native-research-artifact/research-fuzzer"><img src="https://agentmods.dev/badge/skills/ara-labs/agent-native-research-artifact/research-fuzzer.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 | $0.00150 | $0.02340 |
| Opus 5 | $0.00075 | $0.01170 |
| Sonnet 5 | $0.00030 | $0.00468 |
| Haiku 4.5 | $0.00015 | $0.00234 |
Grade A, and why
research-fuzzer 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 11d 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 — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.
research-fuzzer — the fuzzer's feedback loop, for any investigation
A greybox fuzzer almost never finds the bug on a given run. It still wins, because every single run answers one cheap question: did that reach somewhere new? The answer steers the next input. An agent investigating an open-ended question has no such loop by default: it reacts to its last result, grinds the same corner, and stops when the budget runs out — sampling, not searching.
This skill gives you the fuzzer's loop. The mapping is exact:
| the fuzzer has | you keep (the panel's word) | plain words |
|---|---|---|
| coverage map | explored — settled bets, per sub-question |
where you have already walked |
| seed queue | leads — noticed but never tried |
doors you passed, never opened |
| crash reports | unexplained — outcomes you cannot explain |
the confusion ledger |
| "no new coverage" | novelty — recent results that taught you nothing |
the going-in-circles alarm |
| triage before reporting | gate — before any conclusion |
no claim until the books are clean |
Fuzzing vocabulary stops here: the panel and the rules below use the neutral words, so the skill does not nudge you toward software-shaped experiments when your world is biology, markets, or people.
Nobody — you included — can know what fraction of the world you have covered; that number needs a god's-eye view that does not exist. Every reading above is computed from your own footprints instead. That is the whole trick, and it is the same trick fuzzers use: coverage is always measured against what you have seen, never against all possible behaviors.
Works anywhere — by construction
- Any domain. An "action" is anything that returns information: an experiment, a query, a benchmark run, an interview, a paper read, a grep. Predictions may be quantitative ("loss < 0.5") or qualitative ("most users will cite price"); they only need to be falsifiable.
- Any agent, any harness. This file is the skill. Everything below is
executable by hand with no tooling at all;
scripts/tally.py(stdlib-only Python) is an optional convenience that computes the same panel faster. No network, no packages, no framework. - Any timescale. The notebook is one append-only file at the investigation root — it survives context loss, session restarts, and handoffs to other agents. A new session starts by reading the notebook and printing the panel.
- When not to use it. Single-step lookups and trivial fixes. The loop earns its overhead only when the answer is genuinely unknown and multiple actions will be needed.
- Investigation, not synthesis. This loop is built for probing a world that already exists — why is X happening, what law governs Y, where is the bug. For creative work (designing a system, constructing a proof, writing), apply it only to the investigative episodes inside the work — "will this design choice survive load?" is a bet; the act of creation itself is not.
What ships with it
3 files 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.
- 11d ago First seen · 174 lines · 150 tokens per session scan A 0e2af034392a
research-fuzzer is a skill published in the GitHub repository ARA-Labs/Agent-Native-Research-Artifact (680 stars, last pushed 17d ago), licensed MIT. It adds 150 tokens to every session and 2,340 once invoked, about $0.0007 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 skills, from other repositories
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
local-ai-agents
Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…
next-cache-components-adoption
Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…
insight-error-page
Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…
next-cache-components-optimizer
Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…
next-partial-prefetching-adoption
Turn on Partial Prefetching in a Next.js app and work through the insights it surfaces. Use when the user wants to enable or adopt Partial Prefetching, flip the partialPrefetching flag, opt routes in with export const prefetch = 'partial', audit Link prefetch={true} behavior, preserve existing prefetched UI with…