Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Auto Deep Research
v0.0.1深度研究技能。用于深入调查、验证信息、研究主题。当用户需要调研概念、对比工具、追踪事件、分析趋势时使用。通过拆分问题、多次搜索、多源验证，输出结构化报告。确保每个结论有多个独立来源，不确定点要标注。
⭐ 0· 36·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (deep research) match the implementation: scripts and SKILL.md perform searches, page reads, multi-round aggregation, and produce report files. Required capabilities (network, ability to write output files) are appropriate for this purpose.
Instruction Scope
Instructions explicitly read web pages, call external search/reader APIs, and write outputs to output/{topic-slug}/ (state.json, memo.json, sources.json, report.md). This is expected for a research skill, but it does mean user queries and retrieved page contents are persisted to disk and sent to third-party APIs (Tavily, Jina or DuckDuckGo). SKILL.md does not instruct reading unrelated local files or other system credentials.
Install Mechanism
No install spec; skill is instruction-only with two small shell scripts. No downloads or archives; scripts are simple curl wrappers. This is low-risk from an install perspective.
Credentials
SKILL.md documents optional environment variables (TAVILY_API_KEY, JINA_API_KEY) which are appropriate for the described optional integrations. Registry metadata lists no required env vars — a mild inconsistency (these API keys are optional, not required). No unrelated credentials are requested.
Persistence & Privilege
always:false (no forced presence). The skill writes only to its own output directory under the working dir; it does not request system-wide changes or other skills' configs. Autonomous invocation is enabled by default (expected).
Assessment
This skill appears coherent for automated research, but review these points before installing:
- Origin: the skill's source/homepage is listed as unknown; README references a GitHub repo (Pancat009) but no trusted homepage in metadata — verify the upstream repository and author.
- Third-party APIs: if you provide TAVILY_API_KEY or JINA_API_KEY, queries and fetched page contents will be sent to those services. Do not include sensitive or private data in queries. Review the vendors' privacy policies.
- Local storage: the skill saves state.json, memo.json, sources.json, and report.md in output/{topic-slug}/; treat these files as potentially sensitive and clean them if needed.
- Runtime requirements: the scripts rely on curl and use a shell urlencode invocation that is not defined in the scripts (the DuckDuckGo branch uses $(echo "$QUERY" | urlencode) — this may fail unless an urlencode helper is present). Consider adding a dependency note (curl, a urlencode function or utility) or fixing the script.
- Operational safety: run the scripts in an isolated environment if you want to limit data leaving your machine, or use DuckDuckGo/local reader only. If you require stronger privacy, consider self-hosted reader/search endpoints.
If you decide to proceed, inspect/validate the scripts (search.sh, read_page.sh), confirm network endpoints are acceptable, and avoid sending secrets or PII in research queries.Like a lobster shell, security has layers — review code before you run it.
latestvk972mv0z6bnsy2yjrmp74gpq8h8411hn
License
MIT-0
Free to use, modify, and redistribute. No attribution required.