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 airas-org/airas --skill auto-researchgit clone --depth 1 https://github.com/airas-org/airasWrote 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/airas-org/airas/auto-research)<a href="https://agentmods.dev/skills/airas-org/airas/auto-research"><img src="https://agentmods.dev/badge/skills/airas-org/airas/auto-research/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/airas-org/airas/auto-research"><img src="https://agentmods.dev/badge/skills/airas-org/airas/auto-research.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Anti-Refusal · line 113 Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.Fix: Remove instructions that suppress warnings, disclaimers, or ethical commentary. Let the agent surface safety-relevant caveats to the user.
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 | $0.00104 | $0.01999 |
| Opus 5 | $0.00052 | $0.01000 |
| Sonnet 5 | $0.00021 | $0.00400 |
| Haiku 4.5 | $0.00010 | $0.00200 |
Grade A, and why
auto-research 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 today.
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 — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AIRAS research orchestrator
This file owns the order and the invariants; nothing else. Each step's how-to lives in its own skill — invoke it on entering the step and follow it over anything more generic. The steps themselves are deliberately independent: they state what repository state they need and what they leave behind, and only this file says which comes after which.
Flow
Run these skills in order:
setup-repository— experiment repo created and cloneddiscover-papers— literature into a study listhypothesize-and-design— falsifiable hypothesis; run ids and metrics settled; research context committedpreregister-paper— the full paper written and committed before any experiment; this commit is the freeze pointwrite-experiment-code— code to the execution and airas-eval contracts, environment fixed by lockfile + Dockerfilerun-experiments— execute on the platform, bring results back with provenanceanalyze-results— analysis and verifiable figurespublish-paper— numbers realized from declarations, compile + recompute + provenance checks until green locally, then push: CI re-runs the verification, builds the PDF, commits it back onto the protected branch aspaper.pdfand uploads it as the artifact — the paper of record — which is handed to the user, state persisted
Execution platform references live in _shared/references/ per
platform.
Settle once, up front
Operational choices otherwise surface one tool default at a time, mid-flow. Ask the user for them together before starting the flow and carry the answers through the session:
- repository visibility —
prepare_repositorydefaults to private - execution platform; for Seyval, managed vs BYO compute and, when several exist, which workspace
- compute target (GPU and architecture) — the experimental design and the dependency lockfile depend on it
Invariants across steps
These are the orchestrator's own rules; no step may relax them.
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.
- today Changed · +10 lines b6059399bc24
- 4d ago Changed · +1 lines 0636d9734382
- 5d ago Changed · +74 lines · +4 tokens per session a153884671ee
- 9d ago First seen · 81 lines · 100 tokens per session scan A 5614732144a8
auto-research is a skill published in the GitHub repository airas-org/airas (32 stars, last pushed yesterday), licensed MIT. It adds 104 tokens to every session and 1,999 once invoked, about $0.0005 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
research-proof
Turn vague research ideas, math-heavy claims, AI-lab style agent loops, benchmark claims, causal claims, prototype-readiness claims, design research, prompt-injection-sensitive evidence reviews, and medical-research style questions into falsifiable proof programs with fixed Claim/Verifier/Current…
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.
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.
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.