MiniMax CLI Web Search — 技能工具
v0.1.1Web search via MiniMax MCP using a local CLI wrapper (mcporter), with environment preflight, API-key/config checks, and normalized result formatting. Use whe...
2· 1.3k·2 当前·3 累计
by 安全扫描
OpenClaw
安全
high confidenceThe skill is internally consistent: it is a thin CLI wrapper around a local mcporter client that performs preflight checks and formats web-search results, and it does not request unrelated credentials or perform unexpected system access.
评估建议
This skill is a straightforward wrapper around your local 'mcporter' client. Before installing or running it: (1) confirm the 'mcporter' binary on your machine is from a trusted source and inspect its config (e.g., config/mcporter.json) to verify which remote endpoints and API keys it uses; (2) run the script's --preflight manually to confirm it finds the expected 'minimax' server; (3) avoid pasting API keys when running commands and verify the script's temp-file cleanup behavior in your environ...详细分析 ▾
✓ 用途与能力
The name/description claim web search via MiniMax MCP using a local CLI wrapper (mcporter). The included script calls 'mcporter call minimax.web_search', verifies mcporter and a minimax server in 'mcporter list --json', and formats results. The requested binaries (mcporter, python3) and config paths (mcporter config referenced in docs) match the stated purpose.
✓ 指令范围
SKILL.md instructs only to run the provided wrapper for preflight and queries. The script runs local checks, invokes mcporter, parses JSON with python3, uses mktemp for temporary files and cleans them, and only inspects stderr for auth-related messages. It does not read arbitrary files, exfiltrate data, or call unexpected external endpoints itself (network calls are delegated to mcporter).
✓ 安装机制
There is no install spec (instruction-only), so nothing is downloaded or written by the skill installer. The only runtime dependency is an existing 'mcporter' binary and python3 on PATH, which the script checks for at runtime.
✓ 凭证需求
The skill requests no environment variables or credentials. It relies on the local mcporter configuration for any API keys; the README warns about API-key exposure and advises not printing keys. No unrelated secrets are requested.
✓ 持久化与权限
The skill does not request persistent inclusion (always:false) and does not modify other skills or system-wide agent settings. It only executes a local wrapper script when invoked.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv0.1.12026/2/23
Security hardening: replace fixed /tmp filename with mktemp-based temp files and cleanup via trap.
● 无害
安装命令
点击复制官方npx clawhub@latest install minimax-cli-web-search
镜像加速npx clawhub@latest install minimax-cli-web-search --registry https://cn.longxiaskill.com 镜像可用
本土化适配说明
MiniMax CLI Web Search — 技能工具 安装说明： 安装命令：npx clawhub@latest install minimax-cli-web-search 支持国内镜像加速，使用 --registry https://cn.longxiaskill.com 参数可加速下载
技能文档
Use this skill to run web search through MiniMax MCP from CLI, then return clean, source-first results.
Phase 1: Environment Preparation (must run first)
1) Preflight checks
Run:scripts/minimax_web_search.sh --preflight
This verifies:
mcporterexists- MiniMax MCP server is discoverable (
mcporter list --json) - Server status is healthy (
name=minimax,status=ok)
2) If preflight fails, repair by failure type
mcporter not found
minimax MCP server not ready
config/mcporter.json includes minimax server.
- Verify command/transport is valid.
- Auth/API-key related errors
3) Initiate / smoke test
Run one query after preflight passes:scripts/minimax_web_search.sh --query "latest OpenClaw release" --count 3
If this returns results, environment is ready.
Phase 2: Search Usage (runtime)
Quick usage
scripts/minimax_web_search.sh --query "your query" --count 5
Supported options
--query: required search query--count: max printed results (default5)--freshness: freshness hint appended to query (optional)--json: normalized JSON output--raw: raw tool JSON output--timeout: command timeout (default35)
Output contract (default text)
- Show top-N results in order
- For each item: title, URL, snippet, date (when available)
- Keep output concise and directly actionable
Agent behavior guideline
- Start with a focused query (3–7 keywords).
- If low quality, rephrase once with narrower terms.
- Return key findings + links (no table required).
- For time-sensitive asks, include time words in query (e.g.,
today,latest, date).
Error model (for reliable automation)
Script exit codes:
0: success2: argument error3: dependency missing (mcporter/python3)4: config/auth issue (MCP server unavailable, API key/auth problems)5: upstream/runtime/network failure6: no results (non-fatal)
Treat code 6 as a normal “no match” outcome, not a crash.
Risks and handling
- CLI/config drift across machines
--preflight before first use in a new environment.
- API key exposure risk
- Temporary file safety
mktemp for stderr/output temp files and cleans them with trap.
- Upstream response variance
--json normalized output for downstream automation.
- Timeout/network instability
--timeout and retry with narrower query/count.
- Weak relevance
Additional reference
- For setup/verification commands and publish readiness checks, read:
references/environment-checklist.mdReference style
When presenting findings, include direct links for verification. Prefer 3–5 high-signal sources over large dumps.