Back to skill

Security audit

热点雷达

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches its trend-reporting purpose, but its HTML reports can execute unsafe browser code from external trend data or an unaudited CDN script.

Install only if you are comfortable with this skill fetching public trend data from third-party endpoints and storing reports locally. Avoid using real account cookies unless you understand they are credentials. Treat generated HTML reports as untrusted content unless the inline JSON serialization and CDN loading issues are fixed or reports are viewed in a contained environment.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (3)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/htmlReporter.js:233
Finding

Stored Script Injection in Generated HTML Reports

Content
View full analysis
const platformData = ` + platformDataJson + `; const crossPlatformData = ` + crossPlatformJson + `; const trendsData = ` + trendsJson + `; const monitorData = ` + monitorJson + `; ``` ### Technical Analysis The values serialized into these JavaScript variables originate from external platform APIs and local snapshot files. Although `JSON.stringify()` produces valid JSON, it does not make the result safe for direct insertion into an HTML script element. In particular, an attacker-controlled string containing an HTML script-closing sequence can terminate the containing script element before the browser's JavaScript parser processes the JSON. For example, a malicious topic can contain a sequence equivalent to: ```html ``` The visible HTML sections use `escapeHtml()`, but that function is not applied to the JSON inserted into the inline ...[truncated 1855 chars]
Remediation
View remediation
` as `\u003e` - `&` as `\u0026` - U+2028 as `\u2028` - U+2029 as `\u2029` 3. Prefer placing safely escaped JSON inside a non-executable element such as: ```html ``` Then parse it using: ```js const reportData = JSON.parse( document.getElementById('report-data').textContent ); ``` 4. Use a maintained serializer specifically designed for embedding data in HTML. 5. Validate all upstream records against a strict schema, including expected scalar types, maximum lengths, numeric ranges, and maximum list sizes. 6. Add a restrictive Content Security Policy. Prefer external, locally bundled scripts and a policy that does not allow arbitrary inline scripts. 7. Add regression tests containing script-closing sequences in every externally controlled string field and verify that opening the generated report cannot execute them. ]]>

T08 · Insecure Dependencies

Warning
Location
scripts/htmlReporter.js:263
Finding

Externally Hosted Executable JavaScript Loaded Without Integrity Verification

Content
View full analysis
``` ### Technical Analysis Each generated report loads and executes Chart.js from a public CDN when the report is opened. The resource is version-pinned, but it is not protected with Subresource Integrity and is not bundled with the audited project. Consequently, the code that executes in the report is not limited to the content reviewed during this audit. A compromise of the CDN, its delivery infrastructure, the published package artifact, or the relevant upstream account could cause modified JavaScript to execute in generated reports. The external request also discloses report-viewing metadata, such as the viewer's IP address, request time, and HTTP headers, to the CDN. ### Attack Path 1. An attacker compromises the CDN delivery path, the relevant hosted artifact, or an upstream publishing account. 2. The attacker replaces the expected Chart.js response with malicious JavaScript. 3. A user opens a generated HTML report while connected to the network. 4. The browser retrieves the script from the external CDN. 5. Because no integrity hash is present, the browser accepts and executes the altered response. 6. The malicious code runs in the report's browser context. ### Impact Assessment The compromised dependency would receive the same browser-side execution privileges as the report's legitimate scripts. It could: - Read and alter all data displayed in the report. - Inject phishing content or malicious navigation. - Initiate network requests from the viewer's browser. - Access data available to the report's origin. - Affect every report viewer who retrieves the compromised resource during the attack window. This issue does not directly grant operating-system privil ...[truncated 131 chars]
Remediation
View remediation
``` 4. Generate the integrity value from a trusted copy of the exact production artifact; do not use an unverified example hash. 5. Apply a Content Security Policy restricting scripts to the approved local origin or the exact required CDN. 6. Consider making reports fully self-contained and functional offline to eliminate runtime supply-chain exposure and viewer tracking by third parties. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/collector.js:59
Finding

Unbounded Buffering of External HTTP Responses

Content
View full analysis
Remediation
View remediation
MAX_RESPONSE_BYTES) { res.destroy(); reject(new Error('Response exceeds maximum allowed size')); return; } chunks.push(chunk); }); res.on('end', function() { try { const data = Buffer.concat(chunks).toString('utf8'); resolve(JSON.parse(data)); } catch (error) { reject(error); } }); ``` 9. Run collection jobs with operating-system or container-level memory limits to reduce the effect of resource-exhaustion failures. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (13)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

该代码片段的实际功能明显窄于声明描述。声明强调“全网热榜追踪器”及多平台榜单聚合、趋势分析、定时推送等核心能力;但代码只是在本地文件系统中维护一个监控配置文件,管理关键词列表和监控启用状态,没有任何网络访问、平台 API 调用、热榜抓取、分析逻辑、推送逻辑或定时任务。虽然“话题监控”中的关键词配置可视为相关子功能,但就该代码块本身而言,其主要目的只是配置管理,不足以支撑声明中的核心产品能力,因此应判定为描述与行为不匹配。

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README instructs users to export authentication cookies for third-party platforms but does not warn that these values are sensitive credentials. Users may paste long-lived cookies into shells, automation systems, logs, or shared environments, which can expose account access or facilitate unauthorized scraping under the user's identity.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The manifest description broadly describes a '全网热榜追踪器' with multiple capabilities but does not define concrete trigger phrases, activation boundaries, or exclusion conditions. For a markdown/manifest file, this can cause unintended invocation because the skill appears applicable to many common requests about trends, monitoring, or reports.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill requests ExecuteCommand even though its stated purpose is web aggregation, reporting, and local file management, none of which inherently require shell execution. Excess command-execution capability increases the blast radius if the skill is later extended unsafely, mis-invoked, or prompted with untrusted input, enabling arbitrary local command execution paths.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill content is entirely written in Chinese and frames the assistant behavior in Chinese without indicating that users may choose another language. This can constitute a locale/language policy issue when no opt-in or language choice is provided.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill describes writing reports and pushing them to Feishu without a prominent user-facing warning about side effects or outbound data transfer. This can lead to accidental disclosure of monitored topics, trend data, or user-provided keywords to external systems, especially in automated or scheduled contexts.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The only configured keyword is the Chinese term "人工智能", which imposes a specific language/locale constraint in the skill configuration. There is no indication in this file that users can opt into this language choice or that the constraint is justified as region-specific behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The natural-language interface and user-facing instructions begin by presenting the skill solely in Chinese, and the command help later continues in the same language. This can violate a language/locale policy when a skill forces one language without giving the user an explicit choice or documenting that the skill is region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

This JavaScript file contains user-facing descriptions and CLI messages entirely in Chinese, beginning with the header and continuing throughout the command help and status output. Under the policy, forcing a specific language without offering a language choice or documenting a justified locale restriction is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The skill stores monitoring configuration and historical ranking data locally but does not clearly warn users about persistent storage. Persistent local data can expose interests, watchlists, and usage history to other local users, backups, or later processes if users are unaware that retention is occurring.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description is written entirely in Chinese and does not indicate any user language choice or locale opt-in. Under the policy criteria, natural-language content that implicitly fixes a language without offering an option can be a locale-policy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

The generated document hard-codes lang="zh-CN", which imposes a specific language/locale on all report consumers. The file also formats dates and UI labels exclusively in Chinese, and there is no natural-language indication that the locale is optional or intentionally limited to a region-specific use case.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The comments at L886-L887 say the function will use injected trendsData to dynamically display trend information, implying visible per-topic trend labels in the generated HTML. In reality, both getTrendLabel and getTrendClass unconditionally return empty strings, so the platform detail sections never show those trends despite the documentation claiming they do.

Content

No source excerpt is available for this finding.