Back to skill
Skillv1.0.7
ClawScan security
stock trading agents · ClawHub's context-aware review of the artifact, metadata, and declared behavior.
Scanner verdict
BenignMar 16, 2026, 3:24 PM
- Verdict
- benign
- Confidence
- high
- Model
- gpt-5-mini
- Summary
- The skill's code, runtime instructions, and required secrets (Tushare API token and an Aliyun/Bailian LLM key) are consistent with a multi-agent stock analysis tool that calls market-data libs and a remote LLM service.
- Guidance
- This package appears to do what it says: it fetches market data (Tushare/AKShare), sends prompts and data to a remote LLM service (DashScope/Aliyun using the ALIYUN_BAILIAN_API_KEY), and writes reports to disk. Before installing: (1) Inspect scripts/config.py to confirm which environment variables are read and how LLM_API_KEY is assembled; (2) treat the provided API keys as sensitive — use keys with minimal privileges and do not reuse high-value platform keys; (3) expect network traffic to Tushare/AKShare endpoints and dashscope.aliyuncs.com; run the skill in an isolated environment or sandbox first; (4) note that setup.py will install many third‑party packages from PyPI — verify dependency versions if you have supply-chain concerns; (5) SKILL.md references BACKGROUND_RULES.md which isn't present in the manifest shown — check for any missing docs. If you need the skill to never send data outside your environment, do not provide the ALIYUN_BAILIAN_API_KEY or run it with a local/no-network configuration.
Review Dimensions
- Purpose & Capability
- okThe name/description (multi-agent stock diagnosis) matches the code and SKILL.md: it fetches market data via Tushare/AKShare, runs multiple agent roles, and calls a remote LLM endpoint for analysis. Required items (TUSHARE_TOKEN, ALIYUN_BAILIAN_API_KEY) are appropriate for those tasks.
- Instruction Scope
- okSKILL.md and the code instruct copying the package into OpenClaw and running Python entrypoints (advisor.diagnose, scripts/*.py). The runtime actions are data collection, LLM calls, report generation, and saving files — all within the stated purpose. I found no instructions to read unrelated system files or to exfiltrate secrets beyond sending prompts/results to the declared external services.
- Install Mechanism
- noteThis is an instruction-and-code skill with a setup.py that declares many third-party dependencies (agentscope, tushare, akshare, openai, requests, etc.). There is no network download/install of arbitrary archives. The dependency list is broad but consistent with the functionality; note that installing will pull these packages from PyPI and run code locally.
- Credentials
- okThe skill declares two secrets: TUSHARE_TOKEN (market data) and ALIYUN_BAILIAN_API_KEY (LLM). Both are justified by the code. No unrelated credentials, config paths, or unexplained secrets are requested. Minor note: setup.py also includes 'openai' in install_requires while the code points LLM client to a DashScope/Aliyun base_url — this is likely harmless but worth checking in scripts/config.py how keys are mapped.
- Persistence & Privilege
- okThe skill does not request always:true and is user-invocable only. It writes reports to disk by design (MD/JSON/PDF) but does not request elevated or cross-skill configuration privileges.