Skillv1.0.2

ClawScan security

中国A股股票智能分析大师 · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

BenignMar 12, 2026, 4:10 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's declared purpose (read‑only A‑share research via a hosted MCP service) matches its instructions and requirements; it is an instruction-only, read-only integration that does not request credentials or install code.
Guidance: This package is internally consistent for a read‑only A‑share research skill, but before enabling it: 1) verify the hosted MCP endpoint configured by your platform is the authentic RedQuant service (do not accept an unknown/untrusted URL); 2) ensure your platform's network policy and logs record outgoing MCP calls so you can audit what is sent to the remote host; 3) note that optional `web_search` can trigger external internet queries — confirm whether your deployment enables that and whether that's acceptable; 4) since the skill can be invoked autonomously by agents, only install it in environments where you trust its configured MCP backend. If you cannot verify the MCP endpoint or platform wiring, restrict or block network access for this skill until verified.

Purpose & Capability: okName/description, tool catalog and workflows consistently describe a read‑only A‑share research skill that uses a remote RedQuant MCP service. There are no unexpected binaries, credentials, or local dependencies declared.
Instruction Scope: noteSKILL.md instructs the agent to call only read‑only MCP tools (the included tool catalog matches). It also allows optional use of `web_search` for public info but restricts this to explicit user requests or when platform data is insufficient — this is reasonable but introduces potential external network calls outside the hosted MCP service.
Install Mechanism: okThere is no install spec and no code files to deploy. As an instruction-only skill, nothing is written to disk and no external packages are fetched by the skill itself.
Credentials: okThe skill requests no environment variables, no credentials, and no config paths. It expects a hosted MCP endpoint to be provided by the platform/operator, which is appropriate for a remote read‑only integration.
Persistence & Privilege: okalways is false and the skill does not request elevated or persistent platform privileges. Autonomous invocation is enabled by default (platform behavior) but not combined with other risky factors here.