Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
A Share Daily Report
v1.0.2生成专业级 A股早报/晚报，包含大盘指数行情、市场情绪、K线走势图、 行业/概念板块排行、个股涨跌榜、主题新闻追踪、综合分析， 输出 Markdown + PNG 图表 + PDF。数据源为东方财富公开 API、 同花顺公开板块页及海外公开行情接口。 Use when asked to create daily...
⭐ 2· 428·5 current·5 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The declared purpose (generate A‑share morning/evening reports) matches the included code. However the project clearly expects API credentials (MX_APIKEY, TUSHARE_TOKEN) and a Feishu open_id for publishing—yet the registry metadata lists no required env vars or credentials. The code also imports and reuses other local skills (tushare-skills, akshare-cn-market, mx-data/mx-search) via absolute workspace paths, which is disproportionate to a standalone report generator and implies dependency on other skills' configuration/data.
Instruction Scope
Runtime instructions and SKILL.md tell the agent to read .env and config/config.yaml, write reports to a base_dir (defaulting to an absolute user path), verify data sources, and optionally publish to Feishu. The code/docs reference reading environment variables and other skills' directories (/Users/yibiao/.openclaw/...), performing network calls to public APIs, and falling back to mock data. The instructions therefore require access to local files and network endpoints beyond simply calling a single external API—raising risk of unintended exposure of other workspace secrets or data.
Install Mechanism
There is no install spec (instruction-only at registry level) and no remote download in the metadata. That lowers install-time risk: nothing in the registry will be fetched from an arbitrary URL. However the bundle includes many Python scripts (data_fetcher, publisher, etc.) which will be executed if the skill is run; the lack of an explicit install does not remove runtime risk.
Credentials
The skill expects/uses multiple sensitive env vars and tokens (MX_APIKEY, TUSHARE_TOKEN, FEISHU_NOTIFY_OPEN_ID) per SKILL.md/README/DATA_SOURCES.md, yet the registry declares none. The skill also references reuse of globally-configured tokens in other skills (e.g., 'tushare-skills' with an inbuilt token) which could cause it to access credentials that belong to other skills. Asking to read a repository-level .env or other skills' code/configs is broader than the stated purpose and increases the risk of secret exposure.
Persistence & Privilege
always:false (no forced permanent inclusion). The skill can be invoked autonomously (default), which is normal for skills. Combined with the above (access to env, other skills, and publishing), autonomous invocation increases blast radius—e.g., the skill could autonomously read .env and publish documents/messages—so run permissions and invocation policies should be considered.
What to consider before installing
Summary of what to check before installing:
- Metadata vs code mismatch: The package documentation expects MX_APIKEY, TUSHARE_TOKEN, and Feishu credentials, but the skill registry did not declare required env vars. Treat the skill as needing API keys and authenticate accordingly.
- Secrets exposure: The skill reads .env and config files and also references other skills by absolute workspace paths. Run it only in an isolated environment (dedicated venv/project) that does not contain unrelated secrets, and inspect your .env before running.
- Inspect publisher code: Review scripts/publisher.py to see exactly which external endpoints it calls and how Feishu tokens are used (token scope, whether tokens are stored or transmitted). Consider removing or disabling the --publish path until you verify it.
- Absolute/hardcoded paths: The default config uses an absolute user path (/Users/yibiao/...). Update config/config.yaml and outdir to safe, project-local paths before running to avoid accidental writes to unexpected locations.
- Dependency imports: The code attempts to import other local skills (tushare-skills, akshare-cn-market). Confirm those imports point to intended, trusted packages in your environment; if they point to shared/global skill directories you don't control, prefer to sandbox or vendor the exact dependencies.
- Run tests and dry-run: Execute the provided tests in an isolated environment (CI or throwaway VM) and run verify_data_sources/verify_data_truth to see which endpoints are contacted and whether mock-mode works.
If you trust the author and will provide the required API keys, run the skill in an isolated workspace and explicitly set only the minimal env vars needed (avoid putting unrelated secrets in the same .env). If you do not control or trust the environment the skill will run in, treat this package as high-risk and do not install it system-wide.Like a lobster shell, security has layers — review code before you run it.
latestvk976p5jsg2nfymbee516q9rwv5843xa6
License
MIT-0
Free to use, modify, and redistribute. No attribution required.