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 yqi96/warranted --skill paper-reproducegit clone --depth 1 https://github.com/yqi96/warrantedWrote 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/yqi96/warranted/paper-reproduce)<a href="https://agentmods.dev/skills/yqi96/warranted/paper-reproduce"><img src="https://agentmods.dev/badge/skills/yqi96/warranted/paper-reproduce/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/yqi96/warranted/paper-reproduce"><img src="https://agentmods.dev/badge/skills/yqi96/warranted/paper-reproduce.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 | $0.00047 | $0.01406 |
| Opus 5 | $0.00023 | $0.00703 |
| Sonnet 5 | $0.00009 | $0.00281 |
| Haiku 4.5 | $0.00005 | $0.00141 |
Grade A, and why
paper-reproduce 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 — 86 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Graph Mapping
The paper's argument is the object under test. Extract it into the graph first; reproduction then tests whether those Grounds hold under independent execution.
paper conclusion -> Claim
paper stated result -> pending Statement(Ground role), source=observed, verification=pending
paper inference from result to conclusion -> Warrant
paper method/standard behind the inference -> Backing
paper acknowledged exception -> Rebuttal
Granularity:
- extract the Claims whose verdicts matter; do not force the whole paper into one giant Claim
- attach each stated result to the specific Claim it bears on
- distinct results, conditions, populations, or scopes are separate pending Grounds
- a paper sub-Claim used as evidence for another Claim -> reference the sub-Claim's id directly in the consuming Warrant's
ground_ids
Obligations
Object-layer work in this channel — implementation, data download, computation, analysis, debugging — is valid only when it extracts, tests, audits, or reconciles a node or Claim status in the reproduction graph. A run not tied to a Ground it verifies or contests is loose work.
| Graph state | Required action |
|---|---|
| A paper conclusion whose verdict matters has no Claim | create_claim, status proposed |
| A paper-stated result must be tested | create_statement(source="observed", verification="pending") |
| A Warrant needs authority | create_statement(...) from the paper's method or standard, then update_node(<warrant>, backing_ids={add:[...]}) |
| The paper acknowledges an exception or limitation | create_statement(rebuttal_for={target_id, target_type}) |
| The initial Claim-Ground-Warrant structure exists | compile_arguments — coherence to test, not proof |
| A pending Ground needs an independent test | delegate execution to code-experimenter |
| Reproduction output mismatches a Ground | route to discrepancy-auditor before any Rebuttal |
| A claimed blocker would halt an obligation | route to discrepancy-auditor before accepting it |
| A result supports its Ground | update_node(verification="verified", attachments=[...]) |
| Reproduction differs but audit finds a setup issue | keep the Ground pending; fix and rerun |
| Audit confirms a real contradiction | create_statement(rebuttal_for=...), then verify that Rebuttal with the audit's own output attached; the Claim moves to disputed/refuted only after that, a passing compile, and evidence assessment |
| A result reveals something the paper never claimed | new observed Statement or Claim; never rewrite the paper Claim |
| A full-scope test looks too slow, large, or costly | do NOT narrow scope for cost or time alone — long-running experiments are expected and accepted; profile via discrepancy-auditor, then delegate the fix to code-optimizer to make the full test feasible |
| Only a narrower scope is verifiable (after optimization is exhausted) | bound the Claim with a qualifier that honestly records the verified scope and degree; the Claim may then be supported within that qualifier — never present the full, unqualified Claim as supported on scoped evidence |
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 · 86 lines · 47 tokens per session scan A df9a0b676eac
paper-reproduce is a skill published in the GitHub repository yqi96/warranted (2 stars, last pushed 9d ago), licensed MIT. It adds 47 tokens to every session and 1,406 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 2026-08-31.
Other skills, from other repositories
proof-checker
A mathematical proof review and repair workflow for LaTeX documents. It checks whether a proof has valid reasoning, addresses identified gaps, reviews the fixes, and produces an audit report.
auto-paper-improvement-loop
An automated editing cycle for a compiled research paper: an external language model reviews it, fixes are applied, and the paper is compiled again.
research-lit
Search and analyze research papers, find related work, summarize key ideas. Use when user says "find papers", "related work", "literature review", "what does this paper say", or needs to understand academic papers.
paper-illustration
A workflow for generating academic illustrations, such as architecture diagrams and method visuals, with image generation and repeated review. Claude plans and checks the figure during the process.
paper-write
A workflow for writing a research paper section by section in LaTeX, a text format used to prepare structured technical documents. It starts from a paper outline and supports several academic venues and citation sources.
experiment-bridge
A workflow that turns an experiment plan into code, reviews the code, runs an initial check, deploys experiments, and collects early results. Deployment means sending the experiment to the environment where it will run, such as a GPU machine.