📦

감사 이력

ai-maestro-code-graph-query - 4 감사들

감사 버전 4

최신 높은 위험

Jun 27, 2026, 04:05 PM

The static analyzer found many repeated patterns, and most weak cryptography matches are false positives from documentation text and generated output. However, the skill contains coercive prompt-control language, requires broad Bash execution, references home-directory executable installation, and makes localhost API calls, so it should not be published without review.

2
스캔된 파일
329
분석된 줄 수
9
발견 사항
codex
감사자

높은 위험 문제 (2)

Prompt Injection Style Mandatory Automation
The skill tells the assistant that graph querying is not optional, must be the default behavior, and must run immediately after reading code. This can override user intent and force repeated command execution during unrelated code review or editing tasks.
Broad External Command Execution Requirement
The skill grants Bash access and repeatedly instructs the agent to run graph shell scripts. Marketplace users would need to trust external scripts on PATH, and command behavior is outside the reviewed skill files.
중간 위험 문제 (2)
Home Directory Executable Installation Reference
The skill instructs users to verify and install scripts under a hidden home-directory path. User-local executable directories can persist across sessions and may shadow trusted commands if PATH order is unsafe.
Localhost API Call Requirement
The skill directs the agent to call a local AI Maestro API endpoint. This is not external exfiltration, but it can disclose local agent state to any service listening on that port.
낮은 위험 문제 (2)
Weak Cryptography Static Matches Are False Positives
The cited lines are documentation, generated marketplace content, or graph command names. No evidence found of cryptographic algorithms, hashing APIs, key handling, or security-sensitive crypto use in the reviewed files.
Generated Report Findings Are Not Executable Skill Behavior
Static findings in output.json describe already generated report metadata and content. They are useful context but do not create additional runtime behavior for the skill.

감지된 패턴

Coercive Agent Control LanguageShell Script Execution from PATH

감사 버전 3

안전

Jan 16, 2026, 02:27 PM

Documentation-only skill with no executable code. All 79 static findings are false positives caused by markdown documentation patterns. The skill instructs AI agents to run local shell scripts from ~/.local/bin/ and make localhost API calls to port 23000. No external network access or sensitive data exfiltration.

3
스캔된 파일
542
분석된 줄 수
1
발견 사항
claude
감사자
보안 문제를 찾지 못했습니다

위험 요인

감사 버전 2

안전

Jan 16, 2026, 02:27 PM

Documentation-only skill with no executable code. All 79 static findings are false positives caused by markdown documentation patterns. The skill instructs AI agents to run local shell scripts from ~/.local/bin/ and make localhost API calls to port 23000. No external network access or sensitive data exfiltration.

3
스캔된 파일
542
분석된 줄 수
1
발견 사항
claude
감사자
보안 문제를 찾지 못했습니다

위험 요인

감사 버전 1

낮은 위험

Jan 10, 2026, 06:49 AM

Prompt-based skill providing behavioral instructions. Executes local scripts from ~/.local/bin/ and makes localhost API calls to the AI Maestro dashboard. No external network access or sensitive data exfiltration patterns detected.

1
스캔된 파일
160
분석된 줄 수
2
발견 사항
claude
감사자
보안 문제를 찾지 못했습니다

위험 요인

⚙️ 외부 명령어 (2)
🌐 네트워크 접근 (1)