中文 EN

Scan Report

Scan New Files

⚠ Low Risk — Risk Score 15/100
Last scan:23 hr ago Rescan
15 /100
trading-coagent
AI 與交易腳本協作技能。AI 每 5 分鐘監控交易訊號，與腳本並存共生
Trading co-agent skill with declared autonomous trading capabilities but implementation is a non-functional skeleton with all trading logic as TODOs - no actual malicious behavior present.
Skill Nametrading-coagent
Duration29.1s
Enginepi
✓
Safe to install
The SKILL.md describes aggressive autonomous trading behavior, but the Python implementation is purely a monitoring scaffold with no real trading execution capability. Before deployment, ensure trading permissions are properly scoped and implement proper safeguards if actual trading logic is added.

Findings 2 items

Severity Finding Location
Low
Missing allowed-tools declaration Doc Mismatch
SKILL.md describes a trading co-agent system but does not declare allowed-tools permissions. For a skill that describes autonomous trading with AI-script collaboration, expected tools like Read, Write, Bash, or skill_invoke should be declared.
Use when: 需要 AI 與自動化交易腳本協作
→ Add an allowed-tools section mapping Bash→shell:WRITE, Read→filesystem:READ, Write→filesystem:WRITE, WebFetch→network:READ if the skill will invoke external scripts.
 SKILL.md:1
Info
Documentation describes unimplemented functionality Doc Mismatch
SKILL.md describes an autonomous trading system capable of executing real trades with position sizing up to 10%. However, scripts/monitor-signals.py contains only TODO stubs with no actual trading execution logic.
# TODO: 實現實際交易邏輯
→ If this is intentional (e.g., AI handles execution while script handles monitoring), clarify this in documentation. Otherwise, this creates a misleading impression of capability.
 scripts/monitor-signals.py:62
ResourceDeclaredInferredStatusEvidence
Filesystem NONE READ ✓ Aligned scripts/monitor-signals.py:96 - opens monitoring_log.jsonl
Network NONE READ ✓ Aligned Scripts mention RPC endpoints but no actual network calls implemented
Environment NONE NONE — No environment variable access in code
Shell NONE NONE — No subprocess or shell execution in code

File Tree

4 files · 15.3 KB · 650 lines
Markdown 3f · 456L Python 1f · 194L
├─ ▾ 📁 references
│ ├─ 📝 failure-analysis.md Markdown 180L · 3.1 KB
│ └─ 📝 trading-checklist.md Markdown 112L · 1.9 KB
├─ ▾ 📁 scripts
│ └─ 🐍 monitor-signals.py Python 194L · 6.1 KB
└─ 📝 SKILL.md Markdown 164L · 4.2 KB

Security Positives

✓ No credential harvesting or API key theft attempts
✓ No data exfiltration or C2 communication
✓ No obfuscation techniques (base64, eval, etc.)
✓ No suspicious network connections or remote script execution
✓ No access to sensitive paths (~/.ssh, ~/.aws, .env)
✓ No subprocess or shell execution present
✓ No persistence mechanisms or cron job setup
✓ Code is well-structured with proper error handling
✓ Async architecture is appropriate for the use case