Back to skill

Security audit

热门标讯挖掘助手-火标网

Security checks for vulnerabilities and agentic risk

Overview

The procurement tools are disclosed and reasonably scoped, but automatic update confirmations leave an important ambiguity about where your API key may be sent.

Ask the publisher to document and pin the authenticated update-confirmation endpoint before allowing that workflow to run automatically. An existing API key skips device registration. Queries reveal your search interests to the provider, some lookups spend credits, and temporary login links grant account access. All bundled files were inspected; external service behavior and data-retention claims were not verified.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code performs local hardware identification rather than procurement-data aggregation or market analysis. Reading and hashing a MAC address is not a disclosed capability or an evident supporting requirement for the declared purpose. The snippet shows neither transmission of the identifier nor any bidding-statistics API calls.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied command performs local device identification rather than tender-data aggregation or market analysis. No connection to the declared functionality is shown, so this is an undeclared capability rather than an evident supporting implementation detail. The snippet does not show transmission or storage of the resulting hash.

Content

No source excerpt is available for this finding.

The credential fallback collects and transmits device characteristics for account registration.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

When credentials are absent, the skill directs the agent to collect platform, architecture, and a MAC-derived hash, transmit them to an account-registration endpoint, and persist the returned credential. Device identification and account provisioning are not direct requirements of the manifest's tender aggregation purpose or its declared API-key dependency. Requiring user consent mitigates the privacy risk but does not make this additional capability part of the stated scope.

Content

No source excerpt is available for this finding.

The skill exposes company-investigation capabilities beyond the manifest's aggregate tender-market analytics.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The manifest at L003 describes industry heat analysis, top purchasers, and aggregated award statistics. L074–L085 instead enable company registry, shareholder, investment, change-history, and project-contact queries; L536–L548 further add annual reports, qualifications, certificates, and credit ratings. These are independent company-investigation workflows rather than necessary implementation details of aggregate market analysis, and some consume paid credits.

Content

No source excerpt is available for this finding.

The skill mandates an account-based upgrade-notification workflow beyond its declared tender-analytics purpose.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

An account-based update check is not inherently unsafe, but this workflow directs the agent to submit an acknowledgement according to response-provided display_ack while reusing its credential and headers. Without an explicit fixed destination or origin check, an attacker-controlled response could redirect that authenticated request and expose the API key.

Content

No source excerpt is available for this finding.

Automatic company lookups transmit user-provided company names without a privacy disclosure.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workflow explicitly instructs the agent to send the company name from the user's request to search_company and continue querying automatically. Unlike a passive API catalog entry, this selects an external operation, but the file provides no user-facing warning that company names and associated research interests will be sent to the 知了标讯 service.

Content

No source excerpt is available for this finding.

The skill includes personal contact retrieval despite its aggregate market-analysis purpose.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The manifest describes industry heat analysis, top-buyer statistics, and aggregated award records. L220–L255 instead document retrieving individual project contacts, including names and full telephone numbers for paid accounts (L237), which is not needed for those aggregate analyses. Masking for some accounts and prohibiting bulk exports do not remove this out-of-purpose capability.

Content

No source excerpt is available for this finding.

Mandatory account checks and update-display acknowledgments extend the skill beyond tender analytics.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

The update workflow instructs the agent to follow response-provided display_ack information using the same credential and request headers. This source does not constrain that destination, creating a potential credential-disclosure path if an attacker can influence the response; the issue is the missing trust boundary, not the mere presence of account checks.

Content

No source excerpt is available for this finding.

The skill mandates an authenticated package-update notification and acknowledgement workflow unrelated to tender-market analytics.

Medium
Category
Excessive Agency
Confidence
88% confidence
Finding

The source permits response-provided display_ack information to guide a POST that reuses the configured credential and headers, without specifying destination validation. If an attacker controls that response field, an agent could send its API key to an unintended endpoint; a free update workflow does not eliminate this risk.

Content

No source excerpt is available for this finding.

The skill exposes paid corporate due-diligence queries outside its stated tender-trend aggregation purpose.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

The manifest describes industry heat analysis, top purchasers, and aggregation of frequent recent tender winners. Instead, L031–L111 provide operational instructions for retrieving company registrations, shareholder histories, investments, annual financial reports, qualifications, and credit ratings; L024 specifies that successful queries consume credits. These detailed company-investigation capabilities are not necessary for the declared aggregate tender analytics, even though the reference limits calls to information requested by users.

Content

No source excerpt is available for this finding.

The skill mandates an account-based upgrade-notification and acknowledgement workflow unrelated to tender analytics.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

L334 requires using the configured credential to query the account balance endpoint at the start of each session specifically to check upgrade notices. L335–L336 then require displaying those notices and submitting an authenticated POST acknowledgement, introducing account access and remote notification-state changes that are not needed for the manifest's industry heat, top-buyer, or award-aggregation analyses.

Content

No source excerpt is available for this finding.

The skill collects and transmits a persistent hardware fingerprint for trial-account registration, outside its stated tender-analytics purpose.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

Lines L065–L099 provide commands to derive a deterministic SHA256 fingerprint from a network adapter's MAC address, and L109–L125 send that fingerprint and platform information to an account-registration endpoint. The manifest describes industry heat analysis and procurement aggregations, not device identification or automated account provisioning. The consent requirement reduces surprise but does not make hardware fingerprinting a direct requirement of the declared analytics capability.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auto-register.md (reported line 130)May include surrounding context.

md
> ### ⚠️ 请求体必须用 JSON 序列化函数生成,不要手拼字符串
>
> 用 `json.dumps(payload)` / `requests.post(url, json=payload)` / `JSON.stringify(payload)`,
> 或 `curl -d @file`;**不要用字符串拼接,也不要用 Python 的 `str(dict)`**
> (后者产出单引号,服务端会报 `Expecting property name enclosed in double quotes`)。
>

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/auto-register.md (reported line 131)May include surrounding context.

md
> ### ⚠️ 请求体必须用 JSON 序列化函数生成,不要手拼字符串
>
> 用 `json.dumps(payload)` / `requests.post(url, json=payload)` / `JSON.stringify(payload)`,
> 或 `curl -d @file`;**不要用字符串拼接,也不要用 Python 的 `str(dict)`**
> (后者产出单引号,服务端会报 `Expecting property name enclosed in double quotes`)。
>
> 历史教训:曾有版本采集 `home_path`,Windows 的 `C:\Users\alice` 直接拼进 JSON 字符串时

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/auto-register.md (reported line 195)May include surrounding context.

text

注意事项:
- 目录不存在时先 `mkdir -p ~/.zlbx`
- 文件已存在时**合并而非覆盖**(保留用户可能的其他配置)
- `source: "auto"` 字段必须写入,**这是后续判断「是否输出自动登录链接」的关键依据**