Token导航 LogoToken导航

Agent Skill · 研究检索

skill-evidenceops(技能证据操作)

skill-evidenceops 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。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。

skill-evidenceops 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为需要联网;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 7,295。

总安装 19,388周安装 816GitHub Stars 1下载 6,789核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.0

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:skill-evidenceops(技能证据操作)
来源仓库:https://github.com/msrovani/skill-evidenceops
安装命令:
openclaw skills install skill-evidenceops
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 skill-evidenceops

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

对图像、视频、音频等媒体进行法医级分类处理。确保证据完整性并通过监管链追踪来源。skill-evidenceops 属于研究检索类 Skill,可作为该场景下的辅助能力补充。适用于司法、取证等专业场景的数据预处理。操作过程需记录哈希值和元数据变更日志。输出不可篡改,建议存储于只读介质中。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install skill-evidenceops;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

skill-evidenceops 怎么安装?

在终端执行 openclaw skills install skill-evidenceops。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

skill-evidenceops 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

skill-evidenceops 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

skill-evidenceops 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

skill-evidenceops 用的人多吗?

公开数据是累计安装 19,388 次,近一周安装 816 次,下载 6,789 次,GitHub 1 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

skill-evidenceops 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
skill-evidenceops

技能证据操作 · 本页

19,3881需要联网MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
evidenceops
description
Forensic media triage with chain of custody. Use when receiving images, videos, audio, PDFs, or documents that need evidence-grade handling, integrity verification, and audit trails.
version
1.0.0
author
OpenClaw Community
metadata
openclaw
emoji
🔬
tools
install
npm
@openclaw/evidence-vault
os
categories
tags

EvidenceOps - Forensic Media Triage

What It Does

EvidenceOps provides forensic-grade handling of media files with complete chain of custody:

  1. Media Intake - Accept images, videos, audio, PDFs, and documents from any channel
  2. Immutable Storage - Store originals in append-only vault with cryptographic hashes
  3. Metadata Extraction - Extract EXIF, file properties, and media information without altering originals
  4. Derivative Generation - Create thumbnails, transcripts, previews in separate folders
  5. Chain of Custody - Maintain tamper-evident audit trail with hash chain
  6. Integrity Verification - Verify evidence hasn't been modified post-ingest
  7. Audit Trail - Complete JSONL audit log for compliance

What It NEVER Does

  • NEVER modifies original evidence files after ingest
  • NEVER stores secrets, API keys, or credentials in manifests or logs
  • NEVER accepts unsanitized paths from user input
  • NEVER executes untrusted code or downloads remote scripts
  • NEVER exfiltrates data to external services without explicit configuration
  • NEVER bypasses channel allowlists or pairing requirements
  • NEVER stores real personal data in example files

Prerequisites

Before using this skill, ensure:

  1. Plugin @openclaw/evidence-vault is installed and initialized
  2. Vault storage directory is configured with appropriate permissions
  3. Channel allowlist is configured for trusted sources only
  4. Retention policies comply with your legal requirements

Workflow

Step 1: Receive Media

When media is received via any channel:

User sends: [image/video/document]

Required Information:

  • File content (from attachment)
  • Original filename
  • Source channel (whatsapp, telegram, email, etc.)
  • Sender identifier
  • Message ID (if available)

Step 2: Create or Select Case

IF user specifies existing caseId:
  USE that caseId
ELSE IF user requests new case:
  CREATE case with format: case-{YYYY}-{NNN}
  EXAMPLE: case-2026-001
ELSE:
  ASK user: "Should I create a new case or add to existing case [case-2026-XXX]?"

Case ID Format: case-{year}-{sequence}

  • Must match pattern: ^case-[a-zA-Z0-9_-]+$
  • Examples: case-2026-001, case-incident-alpha, case-legal-2026-q1

Step 3: Stage Original (Read-Only)

Before ingest:

  1. Save received file to temporary staging area
  2. Calculate SHA-256 hash immediately
  3. Record file size and MIME type
  4. DO NOT modify the file
# Staging directory structure
/tmp/evidence-staging/
├── {caseId}/
│   └── {timestamp}-{filename}

Step 4: Extract Metadata

Extract metadata WITHOUT modifying original:

For Images:

  • EXIF data (camera, GPS, timestamps)
  • Dimensions
  • Color profile

For Videos:

  • Duration
  • Codec information
  • Resolution

For Audio:

  • Duration
  • Sample rate
  • Codec

For PDFs:

  • Page count
  • Author (if embedded)
  • Creation date
# Use Read tool or appropriate extraction commands
# NEVER write back to original file

Step 5: Generate Derivatives (Optional)

Create derivative artifacts in SEPARATE folder:

derivatives/
├── thumbnails/
│   └── {evidenceId}-thumb.jpg
├── transcripts/
│   └── {evidenceId}-transcript.txt
└── previews/
    └── {evidenceId}-preview.pdf

Derivative Types:

  • thumbnail - Reduced resolution image/video preview
  • transcript - Speech-to-text for audio/video
  • preview - PDF or text representation
  • ocr - Extracted text from images

Step 6: Ingest to Vault

Call the evidence.ingest tool:

{
  "filePath": "/path/to/staged/file",
  "filename": "original-filename.jpg",
  "caseId": "case-2026-001",
  "channel": "whatsapp",
  "sender": "user@example.com",
  "messageId": "msg-abc123",
  "retentionDays": 2555,
  "metadata": {
    "exif": { ... },
    "extracted": { ... }
  }
}

Response:

{
  "success": true,
  "evidenceId": "ev-abc123...",
  "sha256": "a1b2c3...",
  "vaultUrl": "file:///vault/cases/case-2026-001/originals/ev-abc123.jpg",
  "timestamp": "2026-02-17T10:30:00.000Z"
}

Step 7: Update Manifest

The manifest is automatically updated by the ingest operation.

Manifest Location: {vault}/cases/{caseId}/manifest.json

Manifest Contents:

  • Case metadata
  • Evidence items with hashes
  • Derivatives
  • Chain of custody entries
  • Retention policy

Step 8: Return Receipt

Provide user with evidence receipt:

📋 EVIDENCE RECEIPT

Case ID: case-2026-001
Evidence ID: ev-abc123...
File: original-filename.jpg
SHA-256: a1b2c3d4e5f6...
Size: 1.2 MB
Received: 2026-02-17 10:30:00 UTC
Vault: file:///vault/cases/case-2026-001/originals/ev-abc123.jpg

✅ Chain of custody established
✅ Original preserved immutably
✅ Audit trail active

Tool Reference

evidence.ingest

Ingest a file into the evidence vault.

Parameters:

ParameterTypeRequiredDescription
filePathstringYesPath to the staged file
filenamestringYesOriginal filename
caseIdstringYesCase identifier
channelstringNoSource channel
senderstringNoSender identifier
messageIdstringNoMessage ID from source
retentionDaysnumberNoRetention period
metadataobjectNoAdditional metadata

Returns: { evidenceId, sha256, vaultUrl, timestamp }

evidence.verify

Verify evidence integrity.

Parameters:

ParameterTypeRequiredDescription
evidenceIdstringYesEvidence to verify
caseIdstringNoLimit search to case

Returns: { verified, details: { originalIntact, hashMatch, lastVerifiedAt } }

evidence.manifest

Get case manifest.

Parameters:

ParameterTypeRequiredDescription
caseIdstringYesCase identifier

Returns: Complete case manifest with all items

evidence.export

Export case as archive.

Parameters:

ParameterTypeRequiredDescription
caseIdstringYesCase identifier
formatstringNo'zip' or 'tar' (default: zip)

Returns: { exportPath, sha256, size, itemCount }

evidence.access_log

Get audit trail.

Parameters:

ParameterTypeRequiredDescription
caseIdstringYesCase identifier
limitnumberNoMax events (default: 100)

Returns: { events: [ ... ], count }

SECURITY POSTURE

Channel Allowlist Configuration

Configure which channels can submit evidence:

# ~/.openclaw/openclaw.json
{
  "plugins": {
    "evidence-vault": {
      "channelAllowlist": ["whatsapp", "telegram", "email"],
      "channelDenylist": ["public-discord"],
      "requirePairing": true
    }
  }
}

Pairing Requirements

For Direct Messages (DMs):

  • REQUIRE pairing before accepting evidence
  • BLOCK unpaired DMs by default
  • LOG rejected ingestion attempts

For Group Channels:

  • VERIFY channel is in allowlist
  • REJECT evidence from untrusted channels
  • NEVER auto-ingest from public channels

Path Sanitization Rules

ALL paths are validated:

  1. No path traversal - ../ sequences rejected
  2. No null bytes - \0 rejected
  3. No home directory - ~/ rejected
  4. Canonicalization - Paths resolved before validation
  5. Base path check - Must be within vault directory

Violations result in: E_PATH_TRAVERSAL error + audit log entry

Destructive Action Confirmation

Before any potentially destructive operation:

  1. EXPORT - Confirm export destination
  2. RETENTION DELETE - Require explicit confirmation (if enabled)
  3. LEGAL HOLD RELEASE - Require explicit confirmation

Confirmation format:

⚠️ DESTRUCTIVE ACTION

You are about to: [describe action]
Case: [caseId]
Items affected: [count]

Type "CONFIRM" to proceed:

Secrets Handling

NEVER include in logs or manifests:

  • API keys
  • Passwords
  • Tokens
  • Credit card numbers
  • Email addresses (redacted)
  • Phone numbers (redacted)
  • SSN (redacted)

Redaction is automatic via regex patterns.

Troubleshooting

Error: E_PATH_TRAVERSAL

Cause: Filename or path contains disallowed characters

Solution:

  • Filenames are automatically sanitized
  • If error persists, check for unusual characters
  • Original filename is preserved in manifest metadata

Error: E_INVALID_MIME

Cause: File type not in allowed list

Solution:

  • Check allowedMimeTypes configuration
  • Add MIME type to allowlist if appropriate
  • Verify file is not corrupted

Error: E_SIZE_LIMIT

Cause: File exceeds maximum size

Solution:

  • Check maxFileSizeBytes configuration
  • Split large files if appropriate
  • Compress before ingest (note in metadata)

Error: E_CHANNEL_BLOCKED

Cause: Evidence from blocked or non-allowlisted channel

Solution:

  • Verify channel in allowlist
  • Check if pairing is required
  • Review security configuration

Error: E_HASH_MISMATCH

Cause: Evidence file modified after ingest

Solution:

  • This indicates potential tampering
  • Review access logs for case
  • Escalate per incident response procedures

Verification Fails

Symptoms: verified: false in verify response

Diagnostic steps:

  1. Check if original file exists
  2. Compare current hash with manifest hash
  3. Review audit log for modifications
  4. Check filesystem permissions

Configuration Example

# openclaw.yaml
plugins:
  evidence-vault:
    driver: filesystem  # or s3
    basePath: /var/evidence-vault
    maxFileSizeBytes: 524288000  # 500MB
    defaultRetentionDays: 2555   # 7 years
    allowedMimeTypes:
      - image/jpeg
      - image/png
      - video/mp4
      - audio/mpeg
      - application/pdf
    channelAllowlist:
      - whatsapp
      - telegram
      - slack
    requirePairing: true
    
    # S3 driver options (if driver: s3)
    s3:
      endpoint: https://s3.example.com
      bucket: evidence-vault
      region: us-east-1
      objectLock: true

Legal and Compliance Notes

LGPD/GDPR Considerations

  • Tag sensitive data with sensitivityTag
  • Configure appropriate retention periods
  • Never store real personal data in examples
  • Implement data subject access request procedures

Retention Policies

SensitivityDefault Retention
public365 days
internal1825 days (5 years)
confidential2555 days (7 years)
restrictedLegal hold

Audit Requirements

All operations logged to JSONL audit file:

  • Case creation/deletion
  • Evidence ingest
  • Verification attempts
  • Export operations
  • Access requests

Audit log retention: Same as case retention

平台分布

OpenClaw

70.07%
按下载量换算4,757

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

未展示

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills