T09 · Insecure Skill Coding Practices
- Location
SKILL.md:54- Finding
Shell Command Injection Through Unsanitized Query Parameters
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This outing-planner skill is not clearly malicious, but it asks the agent to run external CLI commands and automatically create/open a browser report without enough input-safety or user-control safeguards.
Review this skill carefully before installing. Use it only with non-sensitive team details, avoid member names or private medical/family information when possible, confirm any fly.ai data sharing is acceptable, and do not allow automatic file creation/opening unless you trust and sanitize the generated content. Prefer a pinned, local flyai-cli install and safe argument-based command execution.
SKILL.md:54Shell Command Injection Through Unsanitized Query Parameters
html-template.md:91Unsafe HTML Interpolation Followed by Automatic Browser Opening
README.md:191Unpinned Globally Installed Executable Dependency
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<p>为您的团队精心筛选的最佳出行目的地</p>
</div>
<!-- 团队信息卡片 -->
<div class="team-info">
<h2>📋 团队信息汇总</h2>
<div class="info-grid">
The README explicitly describes collecting team preferences and querying an external service for recommendations, but it does not disclose that potentially sensitive group information may be transmitted off-device or that a local HTML report will be generated and opened. In a workplace context, team composition, budgets, special needs, and family-related details can be sensitive, so the missing privacy and file-handling warning creates a real security and privacy risk through uninformed data exposure.
The natural-language content of the skill is entirely in Chinese and does not indicate that users may choose another language or that the skill is intentionally restricted to Chinese-speaking contexts. This can violate language/locale policy when a skill effectively forces a specific language without user opt-in.
The skill instructs the agent to create a local HTML file in the user's home directory and automatically open it, even though destination recommendation does not inherently require filesystem writes or launching local applications. This expands the skill's capabilities from information retrieval into side-effectful local actions, which can surprise users, overwrite files, or be abused as a delivery mechanism for unsafe HTML content.
The instructions require automatic filesystem modification and command execution without warning or obtaining user approval. Even if the generated HTML is benign, silently writing files and opening them violates least surprise and can be leveraged to trigger local browser rendering of untrusted content derived from external data.
The markdown instructions and embedded HTML are written specifically for Chinese output, including lang="zh-CN" and Chinese-only titles/content, which effectively forces a specific language/locale. The file does not mention user opt-in, alternatives, or that the template is intentionally limited to a China-specific use case.
The skill documentation uses Chinese for most instructions and examples, while only briefly including an English subtitle. This effectively forces a language choice on users without stating that the skill is Chinese-only or offering an alternative language option.
This markdown file contains only Chinese-language instructions and examples, with no indication that the skill is China-specific or that users may choose another language. Under the policy, forcing a specific language without opt-in can be a natural-language policy violation.