Back to skill

Security audit

Byted Viking Developer

Security checks for vulnerabilities and agentic risk

Overview

This is mostly a documentation skill, but it repeatedly shows sensitive Viking memory/API traffic using unencrypted HTTP and says it is not intended for OpenClaw.

Install only after reviewing the docs you plan to follow. Prefer HTTPS for all authenticated Viking SDK/API calls, avoid copying HTTP examples into real environments, use least-privilege AK/SK or API keys, pin dependency versions, and take extra care with delete/update examples. Also note that the skill itself says it does not support OpenClaw scenarios.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
resources/VikingDB 向量库/SDK V2参考/Python SDK/安装与client初始化.md:18
Finding

Unpinned and Mutable Third-Party Dependency Installation

Content
View full analysis

Vulnerability Details

File Locations:

  • resources/VikingDB 向量库/SDK V2参考/Python SDK/安装与client初始化.md:18
  • resources/VikingDB 向量库/SDK V2参考/Python SDK/安装与client初始化.md:27
  • resources/VikingDB 向量库/SDK V2参考/Go SDK/安装与client初始化.md:13-16
  • resources/VikingDB 向量库/最佳实践/【向量库】多模态搜索实践(文搜图/图搜图) - API V2.md:62-67

Vulnerability Type: Mutable and unpinned software supply-chain dependencies
Risk Level: Medium

Vulnerable Code

bash
python3 -m pip install -U vikingdb-python-sdk
bash
git clone https://github.com/volcengine/vikingdb-python-sdk.git
bash
go get -u github.com/volcengine/vikingdb-go-sdk
go get -u github.com/volcengine/volcengine-go-sdk
bash
pip3 install --upgrade volcengine
pip3 install --upgrade tos
pip3 install --upgrade requests
pip3 install --upgrade aiohttp
pip3 install --upgrade tqdm

Technical Analysis

The documentation instructs users to install the latest available versions of third-party packages or clone the mutable default branch of a remote repository. No exact versions, immutable commit hashes, package hashes, or lockfiles are specified.

The --upgrade, -U, and -u options make the installed code dependent on package and repository state at execution time. Likewise, cloning a repository without checking out a reviewed commit allows its effective contents to change after this Skill has been audited.

These dependencies appear relevant to the declared Viking SDK integration functionality, and the reviewed instructions do not show intentional dependency confusion, typosquatting, or execution from an unrelated paste service. Nevertheless, the mutable installation process creates a supply-chain exposure: compromise of a package publisher, source repository, transitive dependency, or future release could introduce code that was not present during review.

Attack Path

  1. An attacker compromises a referenced package registry account, ...[truncated 1383 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin every package to an explicitly reviewed version, for example:

    bash
    python3 -m pip install vikingdb-python-sdk==<reviewed-version>
    
  2. For Git-based installation, pin an immutable commit and verify it before installation:

    bash
    git clone https://github.com/volcengine/vikingdb-python-sdk.git
    cd vikingdb-python-sdk
    git checkout <reviewed-commit-sha>
    
  3. Replace Go upgrade commands with reviewed module versions:

    bash
    go get github.com/volcengine/vikingdb-go-sdk@<reviewed-version>
    go get github.com/volcengine/volcengine-go-sdk@<reviewed-version>
    
  4. Remove broad automatic-upgrade flags from reproducible setup instructions.

  5. Publish lockfiles or constraints files and verify package integrity with cryptographic hashes where supported.

  6. Run dependency installation in a dedicated virtual environment or restricted build container without production credentials.

  7. Add automated dependency scanning, provenance verification, and periodic review of pinned releases and transitive dependencies.

  8. Document the expected official package publishers and repositories so users can detect similarly named or substituted packages.

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (279)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The instructions tell users to replace a public endpoint with a PrivateLink endpoint that only supports HTTP, but they do not prominently disclose that this change removes TLS protection. Users may incorrectly assume 'private' also means encrypted, causing credentials, query contents, and knowledge-base data to traverse internal networks unencrypted.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

This endpoint returns highly sensitive conversational data, including original messages and generated memory/event content, yet the sample uses a cleartext HTTP connection. In the context of a developer integration skill, insecure sample code is especially dangerous because users are likely to copy it directly into production or testing systems, causing exposure of private conversation content and metadata in transit.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The specific line configuring scheme="http" causes SDK authentication and API requests to use plaintext transport while handling sensitive cloud credentials. An attacker on the network path could intercept the AK/SK, replay requests, or tamper with responses, leading to account compromise or unauthorized access to memory-store data.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
100% confidence
Finding

This documentation is for developers integrating a memory/knowledge SDK, so insecure sample code is likely to be copied directly into production or internal tools. Because the example sends authenticated API traffic over unencrypted HTTP and omits any warning about credential-transit risk, it materially increases the chance of credential leakage, man-in-the-middle modification, and unauthorized access to memory library data.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The example explicitly sets the request schema to HTTP while sending an HMAC-authenticated API request. Although HMAC protects request integrity, plaintext HTTP exposes request contents and metadata to interception or tampering in transit and can enable replay or misuse if surrounding controls are weak. In an SDK integration guide, developers are likely to copy this code directly, which makes the unsafe transport guidance especially dangerous.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The example explicitly sets scheme="http" while sending authenticated requests and potentially sensitive memory queries, user identifiers, and profile data. Using plaintext HTTP enables interception or modification in transit by a network attacker, which is especially dangerous in a memory/profile retrieval SDK because it can expose personal or behavioral data and authentication material.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The example explicitly configures the SDK with scheme="http" while sending authenticated requests to a memory retrieval service that can include sensitive conversation history, profile data, and API credentials/signatures. Using unencrypted HTTP exposes request metadata, auth material, and retrieved memory contents to interception or modification by any party on the network path, enabling credential theft, data disclosure, and response tampering.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The Python example explicitly configures the client with scheme="http", which encourages transmitting profile memory and authentication material without transport encryption. Because the payload contains user profile data and the API uses authenticated requests, an attacker on the network could intercept or tamper with sensitive data and possibly capture credentials or signed requests.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The document is titled and described as a bulk delete operation, but the sample code creates a filter update task instead. In an SDK integration guide, this mismatch is dangerous because developers may copy the example believing it demonstrates deletion semantics, leading to unintended data modification workflows, incorrect safeguards, or misuse of destructive task APIs in production.

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/控制面SDK/数据集(Collection)/CreateVikingdbCollection.md (reported line 72)May include surrounding context.

| 8192 | 否 | 否 | | doubao-embedding | 240715 | text | 2048
| 512, 1024, 2048 | 4096 | 否 | 是
| | doubao-embedding-large | 240915 | text | 2048 | 512, 1024, 2048, 4096 | 4096 | 否 | 是 | | doubao-embedding-vision | 250328 | text、image及其组合 | 2048 | 2048, 1024
| 8192 | 否 | 是 | | doubao-embedding-vision | 250615 | 兼容241215和250328的用法*。​*另外,支持full_modal_seq(文/图/视频序列) | 2048
| 2048, 1024
| 128k | 否 | 是 |

返回参数

下表仅列出本接口特有的返回参数。更多信息请参见返回结构。

参数类型描述
MessageString有关请求状态的消息。
ResourceIdString新数据集的ID。

请求示例

创建带vector字段的数据集

python
package main

import (
    "fmt"
    "os"

    "github.com/volcengine/volcengine-go-sdk/service/vikingdb"
    "github.com/v

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Python SDK/控制面SDK/数据集(Collection)/createVikingdbCollection.md (reported line 73)May include surrounding context.

| 8192 | 否 | 否 | | doubao-embedding | 240715 | text | 2048
| 512, 1024, 2048 | 4096 | 否 | 是
| | doubao-embedding-large | 240915 | text | 2048 | 512, 1024, 2048, 4096 | 4096 | 否 | 是 | | doubao-embedding-vision | 250328 | text、image及其组合 | 2048 | 2048, 1024
| 8192 | 否 | 是 | | doubao-embedding-vision | 250615 | 兼容241215和250328的用法*。​*另外,支持full_modal_seq(文/图/视频序列) | 2048
| 2048, 1024
| 128k | 否 | 是 |

返回参数

下表仅列出本接口特有的返回参数。更多信息请参见返回结构。

参数类型描述
messagestr请求状态描述,对应 API 字段 Message。
resource_idstr新建数据集 ID,对应 API 字段 ResourceId。

请求示例

创建带vector字段的数据集

python
import os
import volcenginesdkcore
import volcenginesdkvikingdb as vdb
fro

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/Viking 知识库/SDK参考/Go SDK/知识库(Collection)/search_knowledge.md (reported line 18)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/Viking 知识库/SDK参考/Python SDK/知识库(Collection)/search_knowledge.md (reported line 18)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/Viking 知识库/知识库私网连接方式.md (reported line 45)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/Viking 记忆库/最佳实践/图文记忆库最佳实践.md (reported line 14)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/Viking 记忆库/私网连接方式.md (reported line 36)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/控制面SDK/数据集(Collection)/CreateVikingdbCollection.md (reported line 72)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/数据面SDK/embedding.md (reported line 35)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/数据面SDK/检索/检索后处理算子-PostProcess.md (reported line 9)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/数据面SDK/检索/检索后处理算子-PostProcess.md (reported line 122)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/数据面SDK/检索/检索后处理算子-PostProcess.md (reported line 123)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/数据面SDK/检索/检索后处理算子-PostProcess.md (reported line 145)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/数据面SDK/检索/检索后处理算子-PostProcess.md (reported line 153)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/数据面SDK/检索/检索后处理算子-PostProcess.md (reported line 154)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · resources/VikingDB 向量库/SDK V2参考/Go SDK/数据面SDK/检索/检索后处理算子-PostProcess.md (reported line 155)May include surrounding context.

md
检索后处理算子 是 VikingDB 提供的一种**检索后处理能力(PostProcess)**,用于在检索**召回和标量过滤**之后,对候选数据进行进一步的过滤和优化,确保最终返回的搜索结果更加精准、高效。
在 VikingDB 中,后置处理算子可用于:

* **标量权重融合:​**基于业务需求结合数据时间/空间的信息进行融合排序,融合多个自定义的业务指标得分,从而生成一个更贴合业务需求的最终排序分数,确保召回排序结果符合业务预期。
* **结果精准筛选**:基于业务需求通过正则匹配设定筛选规则,如筛选包含某种特征文字信息的商品,确保用户获得有效的检索结果。
* **数据质量提升**:使用正则匹配对检索结果进行格式校验,过滤包含特殊字符、非法内容或广告信息的数据,确保返回内容符合标准,提高数据规范性。
* **检索结果均衡**:对特定字段设置出现频率限制,防止某些值占据过多搜索结果,如热门品牌、热门作者或特定分类过度曝光,保证检索结果的多样性和公平性。