Token导航 LogoToken导航TokenDH.com
开发敏感数据clawhub未标认证来源可访问clear审计通过

reliability-evidence-pack可靠性证据包

Agent Skill

reliability-evidence-pack 用于补充开发相关能力,适合在 OpenClaw 中需要让 Agent 承接开发相关任务时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

10,407

周安装

438

GitHub Stars

公开资料未说明

下载量

3,644
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:reliability-evidence-pack(可靠性证据包)
来源仓库:https://github.com/christineopenclaw/reliability-evidence-pack
安装命令:
openclaw skills install reliability-evidence-pack
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install reliability-evidence-pack

简介

提供工具来使用标准化模式记录、验证和报告代理操作可靠性工件,以实现一致的监控和合规性。

SKILL.md

Reliability Evidence Pack (REP)

A comprehensive runtime system for documenting and verifying agent operational reliability through structured artifact recording, validation, and compliance reporting.

What's Included

Core Scripts (/scripts)

  • rep.mjs - Main CLI for initialization, validation, and bundle management
  • rep-validate.mjs - Schema validation engine for REP artifacts
  • rep-heartbeat-cron.mjs - Records agent heartbeats on a schedule
  • rep-near-miss-cron.mjs - Tracks near-miss reliability events
  • rep-performance-baseline.mjs - Captures performance metrics
  • rep-generate-bundle.mjs - Generates REP bundles from artifacts

CLI Package (/cli)

  • Installable npm package for convenient command-line access
  • See cli/README.md for installation and usage

GitHub Action (/github-action)

  • CI/CD integration for automated validation
  • See github-action/README.md for setup

Schemas (/schemas)

  • JSON Schema definitions for all REP artifact types:

- decision_rejection_log.json - handoff_acceptance_packet.json - memory_reconstruction_audit.json - near_miss_reliability_trailer.json - signed_divergence_violation_record.json

Examples (/examples)

  • Sample artifacts and workflows
  • Integration patterns for different use cases

Required Binaries

  • Node.js (v16 or higher - required for runtime scripts)
  • npm (for CLI package installation)

No other system binaries required.

Installation

Option 1: Direct Scripts

# Clone or copy this bundle to your project
cp -r rep-bundle-v2 /path/to/your/project/rep
cd rep

# Make scripts executable
chmod +x scripts/*.mjs

# Test
node scripts/rep.mjs --help

Option 2: CLI Package

cd cli
npm install -g
rep --help

Option 3: GitHub Action

- uses: ./github-action
  with:
    bundle-path: ./rep

Configuration

Set environment variables to configure behavior:

# Path to artifacts directory (default: ./artifacts)
REP_ARTIFACTS_PATH=./artifacts

# Path to schemas directory (default: ./schemas)  
REP_SCHEMAS_PATH=./schemas

# Log file path (optional)
REP_LOG_FILE=/var/log/rep.log

Usage

Initialize a new REP

node scripts/rep.mjs init --name "my-agent"

Validate artifacts

node scripts/rep-validate.mjs ./artifacts --strict

Run heartbeat recording (cron)

REP_ARTIFACTS_PATH=./artifacts node scripts/rep-heartbeat-cron.mjs

Set up cron (example for crontab)

# Add to crontab - run heartbeat every 5 minutes
*/5 * * * * cd /path/to/rep && REP_ARTIFACTS_PATH=./artifacts node scripts/rep-heartbeat-cron.mjs >> /var/log/rep-heartbeat.log 2>&1

Artifact Types

ArtifactPurpose
agent_heartbeat_recordAgent lifecycle events
decision_rejection_logDecisions and their outcomes
context_snapshotMemory/context state
handoff_acceptance_packetInter-agent handoff validation
near_miss_reliability_trailerNear-miss events
memory_reconstruction_auditMemory integrity checks
signed_divergence_violation_recordPolicy violations

Security Considerations

  • Credentials: This skill does not require or handle credentials
  • File Access: Writes to configured artifacts directory only
  • Cron: Does not modify system crontab - operator must configure
  • Logs: Optional logging to configurable path

License

MIT

Support

  • Documentation: SPEC.md, QUICKSTART.md, INTEGRATION.md
  • Examples: examples/
  • Validation: node scripts/rep-validate.mjs --help

Security Considerations

Sensitive Data

REP captures context snapshots, decision logs, and memory-like artifacts that may contain sensitive information.

  • Set REP_ARTIFACTS_PATH to an isolated, access-controlled directory
  • Review or redact artifacts before sharing externally
  • Consider running in a container or unprivileged account

Signing Keys

The SPEC includes signature fields for artifact integrity, but key management is operator-defined:

  • Do NOT place private keys in the artifacts directory
  • Use external KMS or secure vault for production signing
  • For testing, generate keys externally and pass via environment (future feature)

CI Usage

The GitHub Action is local to your repository:

  • Review github-action/entrypoint.sh before use in public CI
  • Ensure no artifacts leak to external endpoints
  • Use isolated artifact paths in CI environments

Network Behavior

All scripts operate locally:

  • No telemetry or external API calls
  • No automatic updates
  • All file I/O is to configured artifact paths only

Best Practices

  1. Keep artifacts directory separate from source code
  2. Add artifact paths to .gitignore
  3. Rotate logs regularly
  4. Audit artifacts before external sharing

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

98.08%
按下载量换算3,574

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

未展示

权限和风险

敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills