Back to skill

Security audit

Embodied Ai Weekly

Security checks for vulnerabilities and agentic risk

Overview

The skill’s report-generation purpose is coherent, but its publishing workflow can push broad local repository changes to GitHub Pages without a clear final confirmation or tight file scoping.

Install only if you are comfortable with an agent preparing and publishing a public GitHub Pages report. Before use, require a dry run, inspect the generated files and staged diff, stage only the expected report/archive/latest paths, pin or remove the Chart.js CDN dependency, and give the agent narrowly scoped GitHub credentials.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (3)

T03 · Remote Payload Retrieval and Execution

Warning
Location
SKILL.md:188
Finding

Unpinned Remote JavaScript Dependency in Generated Reports

Content
View full analysis
Remediation
View remediation
``` 2. Obtain the integrity value from a trusted source and independently verify it against the reviewed artifact. 3. Prefer bundling a reviewed local copy of Chart.js into the generated report or publication repository, eliminating runtime reliance on mutable remote code. 4. Add a restrictive Content Security Policy that permits scripts only from explicitly approved sources and disallows unexpected script execution. 5. Establish a controlled dependency-update process that reviews and tests each new Chart.js version before updating the pinned reference. ]]>

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:212
Finding

Shell Command Injection Through Unquoted Configuration Placeholders

Content
View full analysis
/.git --depth 1 # 2. 创建当周文件夹 mkdir /-W # 3. 复制文件 cp .html /-W/index.html cp .md /-W/report.md ``` ### Technical Analysis The Skill instructs the Agent to obtain the repository, local workspace path, report path, year, and week from user configuration and substitute those values directly into shell commands. The placeholders are not quoted, escaped, or constrained by documented allowlists. If these commands are constructed as shell strings, metacharacters such as semicolons, command substitutions, redirection operators, or newline characters can alter the intended command. Even benign paths containing spaces, wildcard characters, or leading hyphens can cause incorrect argument parsing. The most security-sensitive value is `` because it is used repeatedly in command positions. However, the repository owner, repository name, report path, year, and week placeholders also require validation. The repository argument additionally begins with attacker-influenced owner and repository components and should not be accepted as an unrestricted shell fragment. ### Attack Path 1. An attacker or untrusted user supplies a crafted configuration value, such as a local path containing a shell operator and an additional command. 2. The Agent replaces `` or another placeholder with the supplied value. 3. The Agent executes the resulting text through a shell. 4. The shell interprets the injected metacharacters rather than treating the complete value as a single path argument. 5. The injected command runs with the operating-system privileges, environment access, filesystem access, an ...[truncated 1214 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:225
Finding

Overbroad Git Staging Can Publish Unrelated or Sensitive Files

Content
View full analysis
Remediation
View remediation
-W/index.html" \ "-W/report.md" \ "archive/index.html" \ "latest/index.html" ``` 2. Validate and safely quote all path components before using them in Git commands. 3. Inspect the staged changes before committing: ```bash git diff --cached --name-status git diff --cached ``` 4. Abort if the staged file list contains anything outside the expected report, archive, and latest-entry paths. 5. Add appropriate `.gitignore` rules for credentials, environment files, temporary output, screenshots not intended for publication, and editor artifacts. 6. Run secret scanning over the staged diff before every push. 7. Consider using a clean temporary clone dedicated to publication and remove it after the workflow completes. 8. Require explicit user approval before pushing changes to the remote default branch. ]]>
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (10)

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

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

Content

Scanner excerpt · references/html_template_guide.md (reported line 72)May include surrounding context.

html
<header>
  <!-- 背景光晕球 -->
  <div class="hero-orb orb-1"></div>
  <div class="hero-orb orb-2"></div>

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

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

Content

Scanner excerpt · references/html_template_guide.md (reported line 256)May include surrounding context.

GitHub 热榜卡片(Top 3)

html
<div class="hot-repos">  <!-- grid-template-columns: repeat(3, 1fr) -->
  <div class="hot-card">
    <div class="hot-rank hr-1">🥇</div>
    <h3>owner/repo-name</h3>

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The templates hard-code Chinese-language output elements such as zh-CN and Chinese titles/content without indicating that locale should be chosen by the user. This can cause unintended publication in the wrong language or locale, which is a consent and usability issue, especially because the workflow culminates in public GitHub Pages publication.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill directs the agent to clone a repository, create/update files, commit, and push to the remote GitHub Pages branch, which changes externally visible content. It does not require an explicit user confirmation immediately before these destructive/publication steps or warn that the action will modify a remote repository and publish content publicly, creating a risk of unintended repository changes or accidental disclosure.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
82% confidence
Finding

The file title and all guidance are written exclusively in Chinese, and the document does not indicate that language selection is optional or limited to a region-specific use case. Under the policy, a fixed language/locale without user opt-in can be a natural-language policy violation.

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 72)May include surrounding context.

1. 新增仓库(过去7天)

text
https://api.github.com/search/repositories?q=<KEYWORDS>+created:>YYYY-MM-DD&sort=stars&order=desc&per_page=10

2. 经典仓库近期更新

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 · references/github_search_guide.md (reported line 9)May include surrounding context.

1. 新增仓库(过去7天)

text
https://api.github.com/search/repositories?q=<KEYWORDS>+created:>YYYY-MM-DD&sort=stars&order=desc&per_page=10

2. 经典仓库近期更新

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 · references/github_search_guide.md (reported line 14)May include surrounding context.

1. 新增仓库(过去7天)

text
https://api.github.com/search/repositories?q=<KEYWORDS>+created:>YYYY-MM-DD&sort=stars&order=desc&per_page=10

2. 经典仓库近期更新

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The template hard-codes lang="zh-CN", and the surrounding document content is also written in Chinese, which signals a fixed language/locale choice. Under the policy, forcing a specific language or locale without user opt-in or a documented justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The guide presents all instructional content in Chinese and does not indicate that users may choose another language. Per the policy, forcing a specific language without user opt-in is a natural-language policy concern unless the locale constraint is justified or optional.

Content

No source excerpt is available for this finding.