synthetic-sciences/openscience is an AI workbench that carries out scientific research by reading papers, forming hypotheses, writing and running code, conducting experiments, analyzing results, and preparing reports. Researchers use it for work in machine learning, biology, physics, and chemistry with remote or local models. Catalogue add-ons extend its scientific workflows through skills and instructions.
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 synthetic-sciences/openscience --skill reproducegit clone --depth 1 https://github.com/synthetic-sciences/openscienceWrote 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/synthetic-sciences/openscience/reproduce)<a href="https://agentmods.dev/skills/synthetic-sciences/openscience/reproduce"><img src="https://agentmods.dev/badge/skills/synthetic-sciences/openscience/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/synthetic-sciences/openscience/reproduce"><img src="https://agentmods.dev/badge/skills/synthetic-sciences/openscience/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 | $0.00116 | $0.01164 |
| Opus 5.5 | $0.00046 | $0.00466 |
| Sonnet 5.5 | $0.00023 | $0.00233 |
| Haiku 4.5 | $0.00012 | $0.00116 |
Grade A, and why
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 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 — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Reproduce
A reproduction is a measurement of a claim. Its value comes from deciding, before anything runs, exactly what number would count as success, and from writing down exactly what was run so the next person can measure it again. A reproduction that adjusts its target after seeing the result is not one.
Non-negotiables
- Freeze the target first. The claim, the metric and its direction, the expected value with its uncertainty, the tolerance, and the dataset, split and conditions it applies to, written down before execution. If the source gives no uncertainty, choose a tolerance and say why.
- Canonical path before substitutes. Run the authors' code, config, checkpoint and data as released. Only when that path is unavailable or broken, build a replacement, and mark the result as a reimplementation.
- Capture everything that determines the number. Code revision (and whether the tree was dirty), data identity and version, environment and package versions, hardware, seeds, the exact command, elapsed time, and every warning.
- Isolate and preserve. New outputs go to a fresh directory; existing outputs are never overwritten. Raw logs stay separate from derived summaries.
- Do not fix the result. A failing reproduction is a finding. Debugging is allowed until the canonical path runs as intended; changing hyperparameters, filters or metrics to get closer to the paper is a new experiment, reported as such.
- One verdict, from the fixed set.
Workflow
- Resolve the exact target and write the admission record.
- Recover inputs: code, data, environment, configuration, checkpoint.
- Run the canonical path (compute_job for anything longer than a few minutes).
- Capture outputs; compute the metric the frozen way; compare with tolerance.
- Verdict and record.
Admission record (write it to reproduction/<name>/target.md):
Claim: <sentence from the source, quoted> Source: <doi/arXiv/section/table>
Metric: <name>, <direction> Expected: <value ± uncertainty or interval> Tolerance: <value, reason>
Data: <name, version, split, filters, license/access> Conditions: <seeds, budget, hardware>
Canonical path: <repo, commit, command> Missing: <what could not be recovered>
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 · 95 lines · 116 tokens per session scan A 7eec1ecbf9be
reproduce is a skill published in the GitHub repository synthetic-sciences/openscience (4,004 stars, last pushed 3d ago), licensed Apache-2.0. It adds 116 tokens to every session and 1,164 once invoked, about $0.0005 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 2026-09-14.
Other skills, from other repositories
tmux-manual-qa
Run a single manual tmux-based QA scenario for the todo continuation feature against the real CLI (./pi-test.sh) in an interactive TUI. Captures scrollback, asserts deterministic pass/fail count markers, and cleans up test fixtures. Use only for the manual-qa milestone features.
audit-harness
Use when auditing HARNESS.md, pre-commit hooks, pre-push hooks, architecture gates, or CI workflows for tunacode-cli. This skill treats any mismatch, skipped gate, or failing check as a critical failure and requires manual one-by-one execution rather than make targets, batch wrappers, or summary-only audits.
knowledge-engineering-quality-and-delivery-performance-security-and-snapshot-gates
A set of quality gates that checks performance, security, dependencies, test coverage, and stable output snapshots.
knowledge-engineering-quality-and-delivery-real-api-qualification-and-e2e
A guide to tests that call real AI providers and verify complete user flows across the CLI, web interface, terminals, and ACP hosts.
tricorder-cli
Full CLI reference for the tricorder daemon (build diagnostics, test results, source lookup, eval comments, logs). This is the fallback invoked by the tricorder skill when the tricorder-mcp MCP tools aren't available or aren't working — invoke tricorder first; it decides whether this is needed.
knowledge-engineering-quality-and-delivery
A project-wide quality and delivery workflow covering tests, release checks, builds, publishing, and bilingual documentation for Blade Code.