← Back to Skills Marketplace
Gate Audit
by
barrett-cryptoDNA
· GitHub ↗
· v1.0.0
· MIT-0
149
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install gate-audit
Description
Use this skill when the user provides AF2, ESMFold, AF3, Boltz-1, SASA, or MD results and needs a Gate 1-4 audit with a Go/Hold/Kill decision.
README (SKILL.md)
Purpose
用于纳米载体疫苗计算设计中的结构、暴露、MD 与终审门禁判定。
本 skill 专门负责 Gate 1–4 的证据整理与 Go / Hold / Kill 判断。
When to use
当用户提供以下任一信息时调用:
- AF2 / ESMFold / AF3 / Boltz-1 预测结果。
- SASA / 表位暴露结果。
- 2–5 ns 平衡审计结果。
- 100 ns MD 结果。
- checkpoint、日志、NPT 密度、RMSD、Rg、关键距离等摘要。
Audit dimensions
必须分开检查以下五项:
- 结构可行性
- 表位可及性
- 动力学稳定性
- 工程可制造性
- 是否值得进入更昂贵计算
Required checks
Gate 1
- 结构拓扑是否合理
- 关键区段是否断裂
- linker 与展示端是否冲突
- 是否存在明显界面错误
Gate 2
- 表位是否动态可及
- 是否长期遮挡
- SASA 是否支持进入 MD
- 是否存在构象塌陷导致暴露丢失
Gate 3
- 2–5 ns 是否达到基本平衡
- 温度、压力、密度是否正常
- 是否存在日志缺失、checkpoint 缺失、输出文件异常
- 是否存在爆炸、持续漂移、异常离子行为
Gate 4
- 100 ns 是否优于负对照或历史基线
- repeat 间是否趋势一致
- 长程下表位是否保持暴露
- 是否具备进入 CpHMD / TI-FEP 资格
Required output format
- 输入结果摘要
- Gate 1 判定
- Gate 2 判定
- Gate 3 判定
- Gate 4 判定
- 主要失败模式
- 决策:Go / Hold / Kill
- 需要补跑的数据
Default decision logic
- 只要存在关键日志缺失、checkpoint 异常或体系健康不明,默认 Hold。
- 若结构逻辑不成立或表位长期不可及,默认 Kill。
- 只有在结构、暴露、动力学三者同时满足时,才 Go。
- CpHMD 和 TI/FEP 仅作为高级终审,不得作为补救一切问题的默认手段。
Warnings
- 不得用单次最佳轨迹替代重复结果。
- 不得只用单个评分指标下结论。
- 不得忽略失败日志。
- 不得把高模型置信度当作物理正确性。
Usage Guidance
This skill is a checklist and will only be useful if you provide the actual AF/ESM/SASA/MD outputs, logs, and checkpoints it expects. It does not install code or request secrets. Before using: (1) ensure you supply complete logs, checkpoints, and baseline/negative-control data so the audit can follow its rules; (2) do not paste sensitive credentials into the inputs; (3) treat its output as an expert-guided summary — have a domain expert review Go/Hold/Kill decisions for critical actions; (4) if you plan to let an autonomous agent use this skill, confirm the agent has only the specific files you intend it to analyze (the skill itself won’t fetch remote data or run simulations).
Capability Analysis
Type: OpenClaw Skill
Name: gate-audit
Version: 1.0.0
The skill bundle contains only metadata and markdown instructions (SKILL.md) for performing a scientific audit of vaccine design results (e.g., AlphaFold, MD simulations). It defines logical criteria for evaluating structural feasibility and epitope accessibility across four 'Gates' without any executable code, network requests, or malicious prompt injection attempts.
Capability Assessment
Purpose & Capability
Name/description ask for audits of AF2/ESMFold/AF3/SASA/MD outputs; the skill requests no extra binaries, env vars, or installs — that is proportionate for a guidance/checklist skill.
Instruction Scope
SKILL.md contains a focused list of audit checks, required outputs, and decision logic; it does not instruct the agent to read unrelated system files, fetch external resources, or exfiltrate credentials. The agent will only act on user-provided prediction/log/MD data.
Install Mechanism
No install spec and no code files are present (instruction-only), so nothing is written to disk and there is no third-party install risk.
Credentials
No environment variables, credentials, or config paths are requested — which is appropriate for a manual-audit guidance skill.
Persistence & Privilege
always is false and disable-model-invocation is default; the skill does not request permanent/global privileges or modify other skills, so persistence/privilege footprint is minimal.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install gate-audit - After installation, invoke the skill by name or use
/gate-audit - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of gate-audit skill for structured vaccine design audit.
- Enables Gate 1–4 audits using structural prediction, exposure, and MD results (AF2, ESMFold, AF3, Boltz-1, SASA, MD, etc.).
- Standardizes evidence gathering and Go/Hold/Kill decision logic.
- Guides users through required checks, failure modes, and data gaps per audit stage.
- Enforces best practices: multiple criteria, multi-repeat validation, and comprehensive logging.
Metadata
Frequently Asked Questions
What is Gate Audit?
Use this skill when the user provides AF2, ESMFold, AF3, Boltz-1, SASA, or MD results and needs a Gate 1-4 audit with a Go/Hold/Kill decision. It is an AI Agent Skill for Claude Code / OpenClaw, with 149 downloads so far.
How do I install Gate Audit?
Run "/install gate-audit" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Gate Audit free?
Yes, Gate Audit is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Gate Audit support?
Gate Audit is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Gate Audit?
It is built and maintained by barrett-cryptoDNA (@barrett-cryptodna); the current version is v1.0.0.
More Skills