Back to skill

Security audit

破晓

Security checks for vulnerabilities and agentic risk

Overview

The skill is not clearly malicious, but it needs Review because it can auto-trigger broad finance workflows, browse third-party sites, save files to a hardcoded Desktop path, and produce explicit trading-style recommendations.

Install only if you want a Chinese A-share market-report workflow that can browse external finance/news sites and create local report files. Before use, narrow the trigger conditions, require confirmation before browsing or saving, change the output directory to a user-approved workspace path, avoid overwrites, and treat any buy/position suggestions as informational rather than personalized financial advice.

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)

T08 · Insecure Dependencies

Warning
Location
reports/daily_report_20260304.html:7
Finding

Unpinned Remote Stylesheets Create a Supply-Chain and Privacy Risk

Content
View full analysis
``` ### Technical Analysis The generated report automatically loads stylesheets from cdnjs and Google Fonts whenever the HTML file is opened with network access. Neither resource has a Subresource Integrity (`integrity`) attribute, and the document does not restrict resource origins with a Content Security Policy. Consequently, the report's rendered content depends on resources that can change after the project has been audited. A compromise of the CDN, its upstream package, the Google Fonts endpoint, or the relevant network trust chain could modify the report's presentation. Malicious CSS could conceal warnings, overlay deceptive links, alter the apparent meaning of financial information, or make attacker-controlled content appear to be part of the report. Loading these resources also discloses connection metadata, including the user's IP address, user agent, access time, and potentially the referring file or page context, to third-party providers. ### Attack Path 1. The Skill generates or distributes the HTML report. 2. A user opens the report in a network-connected browser. 3. The browser automatically requests the two remote stylesheets. 4. A compromised third-party provider or upstream resource returns modified CSS. 5. The modified stylesheet changes the report's presentation, potentially hiding disclaimers or creating deceptive visual elements. 6. The user may rely on the manipulated financial report or interact with content that appears legitimate. ### Impact Assessment The direct impact is limited to users ...[truncated 462 chars]
Remediation
View remediation
``` 5. Generate a self-contained report that does not make network requests when opened. 6. Add an automated build check that rejects unapproved external `href` and `src` origins. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
steps/06-format.md:185
Finding

Hardcoded Out-of-Project Output Path Can Cause Unauthorized File Writes

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
reports/daily_report_20260304.html:448
Finding

Generated Report Contains Plaintext HTTP and Non-Article Outbound Links

Content
View full analysis
阅读全文 → 阅读全文 → 阅读全文 → ``` Representative plaintext HTTP link: ```html 阅读全文 → ``` Additional plaintext HTTP links appear for government, media, and financial sources at lines 658, 714, 790, and 810. ### Technical Analysis The report presents generic homepages as “read full article” destinations, even though `steps/02-fetch-news.md:103-121` explicitly requires complete article URLs containing an article identifier and rejects homepage URLs. This mismatch prevents users from verifying whether the linked source supports the associated headline and summary. Several destinations also use plaintext HTTP. Although `rel="noopener noreferrer"` mitigates reverse-tabnabbing, it does not provide transport encryption or authenticate the destination. A network-positioned attacker can intercept an HTTP navigation and replace the response or redirect the user to a deceptive destination. Because this is a financial report, unverifiable or intercepted links may influence investment decisions or be used as a phishing entry point. ### Attack Path 1. A user opens the generated report and selects a link labeled as the full article. 2. For homepage links, the user is taken to a generic site rather than evidence supporting the report's claim. 3. For an HTTP link, the request traverses the network without transport authentication or encryption. 4. A malicious access point, proxy, ISP-l ...[truncated 829 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • 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
Findings (22)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
77% confidence
Finding

The description presents the skill entirely as a Chinese-language stock report generator and later mandates WeChat-friendly output, but it does not state that this locale is optional or limited to a region-specific compliance need. This can violate language/locale policy when the skill implicitly forces one language without user opt-in.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger keywords include broad terms like '早报', '收盘', '复盘', and '行情分析', which are common in ordinary financial conversation and can cause the skill to auto-load or invoke outside clear user intent. Because the skill is marked auto_load and global, unintended invocation can lead to unnecessary browsing and data retrieval behavior on unrelated prompts.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · SKILL.md (reported line 41)May include surrounding context.

md
必须打开的页面:
1. 东方财富行情中心:https://quote.eastmoney.com/center/gridlist.html#hs_a_board(三大指数、涨跌家数)
2. 东方财富板块排行:https://quote.eastmoney.com/center/boardrank.html(板块涨跌排行)
3. 东方财富资金流向:https://data.eastmoney.com/zjlx/dpzjlx.html(北向资金、主力流向)
4. 东方财富大盘星图:https://quote.eastmoney.com/stockhotmap/(热力图,注意可能遇滑块验证)

⚠️ 数据不准宁可少写,绝不编造!用户一眼就能看出数据对不对!

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · steps/01-realtime-data.md (reported line 81)May include surrounding context.

md
必须打开的页面:
1. 东方财富行情中心:https://quote.eastmoney.com/center/gridlist.html#hs_a_board(三大指数、涨跌家数)
2. 东方财富板块排行:https://quote.eastmoney.com/center/boardrank.html(板块涨跌排行)
3. 东方财富资金流向:https://data.eastmoney.com/zjlx/dpzjlx.html(北向资金、主力流向)
4. 东方财富大盘星图:https://quote.eastmoney.com/stockhotmap/(热力图,注意可能遇滑块验证)

⚠️ 数据不准宁可少写,绝不编造!用户一眼就能看出数据对不对!

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · steps/01-realtime-data.md (reported line 181)May include surrounding context.

md
必须打开的页面:
1. 东方财富行情中心:https://quote.eastmoney.com/center/gridlist.html#hs_a_board(三大指数、涨跌家数)
2. 东方财富板块排行:https://quote.eastmoney.com/center/boardrank.html(板块涨跌排行)
3. 东方财富资金流向:https://data.eastmoney.com/zjlx/dpzjlx.html(北向资金、主力流向)
4. 东方财富大盘星图:https://quote.eastmoney.com/stockhotmap/(热力图,注意可能遇滑块验证)

⚠️ 数据不准宁可少写,绝不编造!用户一眼就能看出数据对不对!

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · steps/04-midday.md (reported line 92)May include surrounding context.

md
必须打开的页面:
1. 东方财富行情中心:https://quote.eastmoney.com/center/gridlist.html#hs_a_board(三大指数、涨跌家数)
2. 东方财富板块排行:https://quote.eastmoney.com/center/boardrank.html(板块涨跌排行)
3. 东方财富资金流向:https://data.eastmoney.com/zjlx/dpzjlx.html(北向资金、主力流向)
4. 东方财富大盘星图:https://quote.eastmoney.com/stockhotmap/(热力图,注意可能遇滑块验证)

⚠️ 数据不准宁可少写,绝不编造!用户一眼就能看出数据对不对!

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The documented trigger scenarios use generic phrases such as '早报', '午评', '收盘', and '股市资讯' without clear disambiguation rules, making accidental activation likely in normal dialogue. In a globally auto-loaded skill, ambiguous triggers increase the chance that the agent will enter a high-action workflow, including external site access, without sufficiently specific user intent.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest emphasizes generating market news reports and investor-friendly output, but this section gives specific directional calls, position sizing, and named securities recommendations. That is a stronger advisory function than a plain news/report generator and materially changes the skill's effective behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest describes a stock-market information generator for investors that produces morning/midday/closing reports to aid decision-making. The '吃瓜现场' section adds workplace/tech gossip and sarcastic commentary unrelated to market reporting, expanding the delivered content beyond the stated purpose.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

This markdown file consistently instructs the workflow in Chinese and explicitly requires extracting a '中文日期', while not offering any user opt-in or alternative locale. Under the policy rule, forcing a specific language or locale without choice is a natural-language policy violation unless the regional constraint is clearly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The entire skill is written as a mandatory Chinese-only workflow, with no indication that users may choose another language or that the locale restriction is an intentional, documented requirement. Under the policy, forcing a specific language without opt-in is a natural-language policy violation unless clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The entire skill is written in Chinese and all required report labels, examples, and operational instructions assume Chinese output and Chinese-market context without any opt-in or alternative language path. This can violate language/locale policy when a skill mandates a specific language absent documented user choice or explicit justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The entire skill is written as mandatory Chinese-language instructions, including output categories and operational guidance, with no indication that another language is allowed or that the user can opt in to this locale. Under the policy, a forced language/locale without user choice is a natural-language policy violation unless clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The file is entirely written as a Chinese-only reporting template and specifies fixed Chinese output structures such as '破晓早报' and Chinese section labels, but it does not provide any opt-in, fallback, or user choice for language/locale. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file title and all required output examples and instructions are exclusively in Chinese, indicating the skill expects a single language output. The policy allows locale constraints only when users can opt in or when the restriction is clearly justified as region-specific, which is not stated here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file specifies the report structure, instructions, and example output exclusively in Chinese, including the required generated report format. Under the policy rule for natural-language violations, forcing a specific language without user opt-in is a locale/language constraint that should be surfaced unless explicitly justified or optional.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file consistently instructs Chinese-language output and uses China-specific locale conventions throughout, including the title, templates, labels, and lang="zh-CN", but does not offer the user any language or locale choice. Under the policy, forcing a specific language/locale without opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This step goes beyond simple formatting and directs the agent to save files locally and export to PDF using browser automation or external tooling. In an agent context, instructing filesystem writes and tool-driven document conversion increases the capability surface and can cause unprompted local side effects, especially when the user only asked for formatted output.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill instructs writing generated reports to a fixed desktop path under a specific home directory without explicit user confirmation. Hardcoded local writes can leak sensitive report contents, overwrite existing files, fail unpredictably across environments, and violate least surprise when a user expected only chat output.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

The HTML root explicitly sets lang="zh-CN", and the entire report content is presented only in Simplified Chinese. Under the stated policy, forcing a specific language or locale without user opt-in is a natural-language policy concern unless the locale restriction is clearly justified.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

This markdown file instructs the agent to open Eastmoney pages and scrape live market data, which necessarily involves outbound network requests to a third-party site. The instructions are mandatory and detailed, but the description does not include any warning that the skill will contact external services or transmit browsing metadata/user system network traffic.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

Generating market briefings and formatting them for WeChat does not inherently require shell-based document conversion utilities. Invoking external command-line tooling introduces an execution-oriented capability that is outside the core, explicitly stated purpose of generating formatted stock reports.

Content

No source excerpt is available for this finding.