Low Risk — Risk Score 25/100
Last scan:19 hr ago Rescan
25 /100
daily-backup
每日 Git 备份。提交工作区所有变更,记录变更摘要。触发时机:cron 定时任务或手动调用。
Documentation-only skill with declared-permission mismatch but no actual executable code or scripts present to evaluate for malicious behavior.
Skill Namedaily-backup
Duration35.6s
Enginepi
Safe to install
Add missing permission declarations in SKILL.md YAML frontmatter for filesystem:READ (spec reading), filesystem:WRITE (report generation), shell:WRITE (script execution), and network:WRITE (Feishu API). Create actual scripts/auto-backup.sh with documented, pinned git operations.

Findings 2 items

Severity Finding Location
Low
Undeclared permission requirements Doc Mismatch
SKILL.md YAML frontmatter declares 'allowed-tools: NONE' but the documented behavior requires filesystem:READ, filesystem:WRITE, shell:WRITE, and network:WRITE to function as described.
allowed-tools: NONE
→ Declare required permissions in YAML frontmatter matching the documented workflow.
SKILL.md:1
Low
Referenced script does not exist Doc Mismatch
The skill references scripts/auto-backup.sh for core functionality but no such file exists in the package. Without the actual implementation, the true behavior cannot be audited.
运行 `scripts/auto-backup.sh`
→ Include the actual backup script implementation for security review.
references/spec.md:6
ResourceDeclaredInferredStatusEvidence
Filesystem NONE READ ✗ Violation SKILL.md:8 - reads references/spec.md
Filesystem NONE WRITE ✗ Violation SKILL.md:17 - writes data/exec-logs/daily-backup/ reports
Shell NONE WRITE ✗ Violation references/spec.md:6 - runs scripts/auto-backup.sh
Network NONE WRITE ✗ Violation SKILL.md:14 - sends reports to Feishu

File Tree

2 files · 1.7 KB · 78 lines
Markdown 2f · 78L
├─ 📁 references
│ └─ 📝 spec.md Markdown 42L · 764 B
└─ 📝 SKILL.md Markdown 36L · 955 B

Security Positives

✓ No malicious code present - only documentation files exist
✓ No sensitive file access patterns detected
✓ No network exfiltration indicators
✓ No obfuscation or base64-encoded payloads
✓ No credential harvesting mechanisms
✓ No suspicious dependencies or supply chain risks