技能 agentic-trust

📦

agentic-trust

Name: agentic-trust
Author: neo-daniil

內容修訂 r2 高風險 ⚙️ 外部命令🌐 網路存取🔑 環境變數

使用 Agentic Trust 評估服務

選擇代理執行服務需要一致的證據與評價歷史。此技能會引導目錄搜尋、證據檢查、問卷載入，以及結構化評價提交。

支援: Claude Codex Code(CC)

⚠️ 38 差

Maintain this Skill? Claim it ⚠️ 38 差

安裝前請停止並請求確認。

請審查安裝計畫，並在變更檔案前取得使用者明確同意。

使用我的智慧代理安裝

將此請求複製給你的智慧代理，其中包含權威 Skill 頁面和清單。

智能體請求

Review the Skillstore skill "agentic-trust" from https://skillstore.io/skills/neo-daniil-agentic-trust.md and its manifest at https://skillstore.io/api/skills/neo-daniil-agentic-trust/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

你的智慧代理仍應顯示其計畫，並依安全政策請求必要確認。

Agent 可讀資源

當 AI Agent、爬蟲或腳本需要乾淨脈絡、而不是讀取完整頁面時，請使用這些連結。

Markdown 詳情 GET /skills/neo-daniil-agentic-trust.md 已簽名 manifest GET /api/skills/neo-daniil-agentic-trust/manifest 已簽名 lockfile GET /api/skills/neo-daniil-agentic-trust/lockfile

測試它

正在使用「agentic-trust」。搜尋付款工作流程服務。

預期結果:

包含服務名稱、公開分數、評價數量、證據註記與信心注意事項的排名摘要。

正在使用「agentic-trust」。檢查特定服務與記憶檔案。

預期結果:

一份簡潔報告，區分服務事實、評價趨勢、本機既有評分與建議。

正在使用「agentic-trust」。提交任務後評價。

預期結果:

顯示已接受評價、彙總變更，以及本機記憶是否已更新的確認訊息。

安全審計

高風險

v6 • 2026/8/9 開啟版本化報告

Markdown backticks and service-oriented inspection text are false positives, while the Agentic Trust URLs represent intentional network access. The audit confirms unsafe command-line API-key handling and identifies unconfirmed external publication plus an unpinned, missing CLI dependency.

已掃描檔案

224

分析行數

複核項目

已忽略的誤報

已確認的安全問題 (2)

高

SKILL.md:99-111

External Review Publication Lacks Explicit Confirmation

The default write flow submits scores, public text, and a note with publication consent already approved, but it does not require immediate user confirmation.

The submit-review example clearly performs an authenticated write and sets publish consent to approved. No confirmation step appears in the surrounding write workflow.

中

SKILL.md:40-44 SKILL.md:180-195

Unpinned Remote Executable Dependency

The package references a remotely hosted CLI and instructs agents to use a local script that is absent from the scanned package, without version or integrity pinning.

The skill names the remote CLI URL and repeatedly invokes scripts/trust_catalog_cli.py, while the scanned package contains only SKILL.md and LICENSE.txt.

能力複核項目 (17)

這些是真實的本機能力，對此技能而言可能是預期行為，因此需要複核，但不會按已確認的惡意行為計分。

高

SKILL.md:101

Generic API/secret keys

--api-key "$API_KEY" \

The example expands API_KEY into a command-line argument for an authenticated external request. Expanded secrets can be exposed through process inspection and unverified CLI handling.

低

SKILL.md:33

Hardcoded URL

- Base URL: `https://agentictrust.top`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:34

Hardcoded URL

- Hosted skill: `https://agentictrust.top/skills/agentic-trust/SKILL.md`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:35

Hardcoded URL

- OpenAPI JSON: `https://agentictrust.top/openapi.json`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:36

Hardcoded URL

- Swagger UI: `https://agentictrust.top/v1/docs`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:37

Hardcoded URL

- Questionnaire: `https://agentictrust.top/v1/questionnaire`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:38

Hardcoded URL

- Public catalog: `https://agentictrust.top/catalog`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:41

Hardcoded URL

- CLI: `https://agentictrust.top/skills/agentic-trust/scripts/trust_catalog_cli.py`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:42

Hardcoded URL

- Payload reference: `https://agentictrust.top/skills/agentic-trust/references/payloads.md`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:43

Hardcoded URL

- Answers template: `https://agentictrust.top/skills/agentic-trust/references/answers-template.json`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:44

Hardcoded URL

- Review memory template: `https://agentictrust.top/skills/agentic-trust/references/review-memory-te

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

低

SKILL.md:68

Hardcoded URL

python3 scripts/trust_catalog_cli.py discover --base-url https://agentictrust.top --q "payments" --l

The documented CLI command explicitly targets agentictrust.top, so executing it performs intentional external network access. The destination is visible and fixed in the example.

低

SKILL.md:71

Hardcoded URL

--base-url https://agentictrust.top \

The documented CLI command explicitly targets agentictrust.top, so executing it performs intentional external network access. The destination is visible and fixed in the example.

低

SKILL.md:75

Hardcoded URL

python3 scripts/trust_catalog_cli.py ranking --base-url https://agentictrust.top --kind top --limit

The documented CLI command explicitly targets agentictrust.top, so executing it performs intentional external network access. The destination is visible and fixed in the example.

低

SKILL.md:93

Hardcoded URL

python3 scripts/trust_catalog_cli.py questionnaire --base-url https://agentictrust.top

The documented CLI command explicitly targets agentictrust.top, so executing it performs intentional external network access. The destination is visible and fixed in the example.

低

SKILL.md:100

Hardcoded URL

--base-url https://agentictrust.top \

The documented CLI command explicitly targets agentictrust.top, so executing it performs intentional external network access. The destination is visible and fixed in the example.

低

SKILL.md:201

Hardcoded URL

- raw URL: `https://agentictrust.top/skills/agentic-trust/references/payloads.md`

This canonical link directs the agent to content or APIs on agentictrust.top, creating an intentional external network dependency. The link alone does not transmit credentials.

風險因素

⚙️ 外部命令 (50)

🌐 網路存取 (16)

SKILL.md:33 SKILL.md:34 SKILL.md:35 SKILL.md:36 SKILL.md:37 SKILL.md:38 SKILL.md:41 SKILL.md:42 SKILL.md:43 SKILL.md:44 SKILL.md:68 SKILL.md:71 SKILL.md:75 SKILL.md:93 SKILL.md:100 SKILL.md:201

🔑 環境變數 (1)

SKILL.md:101

審計單位: codex 檢視審計紀錄 →

分享與引用此報告

分享版本化評估報告、中性徽章、嵌入卡片與引用。Skillstore 只報告證據，不判斷該 Skill 是否安全。

開啟版本化報告

複製報告連結

https://skillstore.io/skills/neo-daniil-agentic-trust/audits/6?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown 徽章

[![Skillstore security assessment](https://skillstore.io/badges/skills/neo-daniil-agentic-trust/security.svg)](https://skillstore.io/skills/neo-daniil-agentic-trust?utm_source=security_passport_badge)

HTML 徽章

<a href="https://skillstore.io/skills/neo-daniil-agentic-trust?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/neo-daniil-agentic-trust/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

嵌入卡片

<iframe src="https://skillstore.io/embed/skills/neo-daniil-agentic-trust.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

學術引用 (APA · BibTeX · CFF)

APA 引用

neo-daniil. (2026). agentic-trust security audit report (audit version 6) [Author version unspecified]. Skillstore. https://skillstore.io/skills/neo-daniil-agentic-trust/audits/6

BibTeX 引用

@techreport{neo-daniil-neo-daniil-agentic-trust-2026,
  author = {neo-daniil},
  title = {agentic-trust security audit report (audit version 6)},
  institution = {Skillstore},
  year = {2026},
  number = {6},
  url = {https://skillstore.io/skills/neo-daniil-agentic-trust/audits/6},
  note = {Author version unspecified}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "agentic-trust security audit report (audit version 6)"
version: "unspecified"
type: report
authors:
  - name: "neo-daniil"
date-released: "2026-08-09"
url: "https://skillstore.io/skills/neo-daniil-agentic-trust/audits/6"
identifiers:
  - type: other
    value: "skillstore:neo-daniil-agentic-trust:audit:6"
    description: "Skillstore immutable audit report identifier"

Skillstore 評分

評分依據證據置信度: 中

架構

可維護性

內容

社群

規範符合度

你可以打造什麼

比較服務選項

在選擇服務前檢視公開分數、證據與本機歷史。

提交任務評價

使用問卷答案與唯一 fingerprint 記錄已完成的任務。

稽核信任證據

在接受供應商前檢查已發布的評價與問卷結構。

試試這些提示

尋找可信任服務

使用 Agentic Trust 搜尋符合 [need] 的服務。顯示服務名稱、分數、評價數量與關鍵證據。

檢查單一服務

使用我位於 [path] 的記憶檔案檢查服務 [service_id]。將觀察到的事實與結論分開呈現。

準備評價

擷取 [category] 目前有效的問卷。比較 [service_id] 的既有記憶，然後草擬整數答案與理由。

提交並記錄評價

使用 [answers_file]、checksum [checksum]、任務 fingerprint [fingerprint]，以及我提供的 API 金鑰，為 [service_id] 提交評價。接受後更新記憶。

最佳實務

每次寫入操作前都先擷取即時問卷。
為每個真實評價情境使用唯一的任務 fingerprint。
記錄新分數與先前本機評分不同的原因。

避免

請勿重複使用任務 fingerprint 來建立重複評價。
請勿將 API 金鑰傳送至不受信任的 base URL。
未經使用者核准，請勿發布評價文字。

常見問題

此技能需要網際網路存取權嗎？

是。目錄搜尋、服務檢查、問卷載入與評價提交都會呼叫 Agentic Trust 相容 API。

它可以提交評價嗎？

可以。當你提供 API 金鑰、服務 ID、答案檔案與任務 fingerprint 時，它可以提交結構化評價。

它會在本機計算信任分數嗎？

不會。伺服器會根據提交的答案計算指標分數與彙總信任訊號。

什麼是本機評價記憶？

它是一個 JSON 檔案，用於儲存先前已接受的評分，讓之後的評價能保持一致。

公開評價文字是選用的嗎？

是。只有在發布同意獲准時，才應使用公開文字。

使用寫入模式前應驗證哪些項目？

請驗證 base URL、問卷 checksum、服務 ID、API 金鑰來源與任務 fingerprint。

開發者詳細資訊

作者

neo-daniil

授權

MIT

Skillstore 修訂

版本提示

作者未聲明版本。

儲存庫

https://github.com/neo-daniil/agentic-trust-skill/tree/main/

引用

02be9409c79ca1183f7844009c14d9df684d0cf9

維護新鮮度

2026/8/11

使用量

9 次下載 · 125 次瀏覽

檔案結構

📄 LICENSE.txt

📄 SKILL.md