← 返回 Skills 市场
CertainLogic Verifier - Hallucination Guard
作者
CertainLogicAI
· GitHub ↗
· v1.0.0
· MIT-0
73
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install certainlogic-verifier
功能描述
Install, configure, and use CertainLogic Verifier (hallucination‑guard) – deterministic AI verification middleware that catches hallucinations before they re...
安全使用建议
This package appears coherent with a self‑hosted hallucination‑guard: it clones a GitHub repo, installs Python deps, and runs a local HTTP service that your agents call. Before installing: (1) inspect the repository (requirements.txt, Dockerfile, main app) and verify the GitHub project exists and is trustworthy; (2) avoid piping unknown curl output into bash — clone and review first; (3) run in an isolated environment (container or VM) and limit network access if you want an air‑gapped setup; (4) do not provide sensitive credentials unless you need the OpenRouter fallback (OPENROUTER_API_KEY) and trust that service; (5) replace default passwords (e.g., postgres changeme) and audit the audit_log.jsonl storage/rotation. If you want higher assurance, ask the maintainer for a signed release or a reproducible build and confirm the repo URL/ownership before deployment.
功能分析
Type: OpenClaw Skill
Name: certainlogic-verifier
Version: 1.0.0
The skill bundle provides instructions and scripts for installing and configuring 'CertainLogic Verifier,' a self-hosted middleware for AI response validation and semantic caching. The installation script (scripts/install.sh) and SKILL.md follow standard procedures for setting up a Python-based service, including repository cloning and virtual environment management. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found; the requested environment variables (e.g., OPENROUTER_API_KEY) are consistent with the tool's documented functionality for LLM fallback.
能力标签
能力评估
Purpose & Capability
Name/description match the provided artifacts: SKILL.md, API docs, sample facts, integration guides, docker-compose, and an installer script all describe a local verifier service that validates LLM responses against a facts DB and provides caching/audit logs. Required resources and integrations (local HTTP endpoints, optional OpenRouter fallback) are coherent with that purpose.
Instruction Scope
Instructions are narrowly scoped to installing and running a local web service, populating a facts DB, and calling local endpoints. They do not request access to unrelated system files or credentials. One minor scope note: the docs advertise a one‑line installer (curl | bash) pattern in the install.sh comment and provide a script that clones and pip‑installs the repo — this is normal for self‑hosted projects but gives the operator discretion to run arbitrary code from the referenced repository, so inspect before executing.
Install Mechanism
No platform installer declared in registry (instruction‑only), but the SKILL.md and scripts instruct git‑cloning a GitHub repo and running pip install -r requirements.txt (and include docker/docker‑compose options). GitHub is a common release host, so this is expected, but installing packages from an unreviewed requirements.txt and running the repo code executes remote code — review requirements.txt, Dockerfiles, and main application code before installing in production.
Credentials
Registry metadata listed no required env vars, but the documentation references a small set of environment variables (PRODUCT_MODE, LOG_LEVEL, CACHE_DIR) and an optional OPENROUTER_API_KEY used only for cache‑miss warmup/fallback. That key is proportional to the advertised 'warm‑up using OpenRouter' feature, but it is sensitive — only provide it if you intend to enable that functionality and run the service in a trusted network. Docker examples include a placeholder DB password ('changeme') — replace defaults in production.
Persistence & Privilege
The skill does not request always:true, does not require modifying other skills, and does not declare persistent platform‑level privileges. It is a self‑hosted service the operator runs; nothing in the package auto‑enables itself across other agent configs.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install certainlogic-verifier - 安装完成后，直接呼叫该 Skill 的名称或使用
/certainlogic-verifier触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
Initial public release: self-hosted deterministic AI verification middleware with semantic caching, 98% token savings, cryptographic audit logs, and 99%+ hallucination block rate. pip-installable. Works with any LLM pipeline.
元数据
常见问题
CertainLogic Verifier - Hallucination Guard 是什么？
Install, configure, and use CertainLogic Verifier (hallucination‑guard) – deterministic AI verification middleware that catches hallucinations before they re... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 73 次。
如何安装 CertainLogic Verifier - Hallucination Guard？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install certainlogic-verifier」即可一键安装，无需额外配置。
CertainLogic Verifier - Hallucination Guard 是免费的吗？
是的，CertainLogic Verifier - Hallucination Guard 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
CertainLogic Verifier - Hallucination Guard 支持哪些平台？
CertainLogic Verifier - Hallucination Guard 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 CertainLogic Verifier - Hallucination Guard？
由 CertainLogicAI（@certainlogicai）开发并维护，当前版本 v1.0.0。
推荐 Skills