Trust Assessment
game-marketing received a trust score of 65/100, placing it in the Caution category. This skill has some security considerations that users should review before deployment.
SkillShield's automated analysis identified 4 findings: 0 critical, 2 high, 2 medium, and 0 low severity. Key findings include Excessive Permissions: Broad API/DB Access for Data Extraction, Excessive Permissions: Automated Public Channel Posting, Excessive Permissions: File System Write Access.
The analysis covered 4 layers: Manifest Analysis, Static Code Analysis, Dependency Graph, LLM Behavioral Safety. The LLM Behavioral Safety layer scored lowest at 56/100, indicating areas for improvement.
Last analyzed on February 13, 2026 (commit 13146e6a). SkillShield performs automated 4-layer security analysis on AI skills and MCP servers.
Layer Breakdown
Behavioral Risk Signals
Security Findings4
| Severity | Finding | Layer | Location | |
|---|---|---|---|---|
| HIGH | Excessive Permissions: Broad API/DB Access for Data Extraction The skill describes automated tasks that require the AI agent to extract data from 'game API/DB', 'bot API / analytics', and 'ad network dashboards'. Granting unconstrained access to these systems could allow the agent to read and potentially exfiltrate sensitive user data, game statistics, or financial information without proper authorization or logging. This broad access poses a significant data exfiltration risk. Implement fine-grained access controls for all external APIs and databases. Ensure the agent only has read access to necessary fields and that all data extraction is logged and requires explicit user confirmation for sensitive operations. Consider using a proxy or middleware to filter and validate data requests. | LLM | SKILL.md:241 | |
| HIGH | Excessive Permissions: Automated Public Channel Posting The skill outlines tasks such as '텔레그램 채널에 자동 전송' (auto-post to Telegram channel) and '자동 포스팅' (auto-post) for social media content (TikTok, Reddit, X, blog). Allowing an AI agent to automatically post to public channels without strict human review and approval for each post carries a high risk of publishing inappropriate content, spam, or misinformation, which could damage brand reputation, lead to account suspension, or be used for phishing/scamming. Require explicit human approval for all public posts. Implement a review queue where the agent drafts content, but a human must approve it before publishing. Ensure the agent's posting capabilities are strictly scoped to designated accounts and channels. | LLM | SKILL.md:244 | |
| MEDIUM | Excessive Permissions: File System Write Access The skill mentions '포스트 초안 작성 → _drafts/ 폴더에 저장' (Draft post -> save to _drafts/ folder). This implies the agent would have write access to the local file system. Unrestricted write access could allow an attacker (via prompt injection or other means) to write malicious files, overwrite critical system files, or consume disk space, potentially leading to denial of service or further compromise. Restrict file system write access to a dedicated, sandboxed directory. Implement strict validation on file names and content. Ensure the agent cannot write to system-critical paths or execute arbitrary files it creates. | LLM | SKILL.md:290 | |
| MEDIUM | Command Injection Risk: External Tool Invocation (MiniPC Playwright) The skill mentions '게임플레이 스크린샷/GIF 자동 캡처 (MiniPC Playwright)'. This suggests the AI agent might invoke an external tool or library named 'MiniPC Playwright'. If this tool is executed via a shell command and its arguments can be influenced by untrusted input (e.g., user-provided game names or paths), it could lead to command injection, allowing an attacker to execute arbitrary commands on the host system. Even without direct injection, invoking external tools without strict sandboxing and input validation is an excessive permission that could be abused. Avoid direct shell execution of external tools. If necessary, use a secure API or library wrapper. Ensure all inputs to external tool invocations are strictly validated and sanitized. Run the agent in a highly sandboxed environment with minimal privileges. | LLM | SKILL.md:262 |
Scan History
Embed Code
[](https://skillshield.io/report/59c557f90cad470c)
Powered by SkillShield