ClawHub

MiniMax CLI Web Search

Web search via MiniMax MCP using a local CLI wrapper (mcporter), with environment preflight, API-key/config checks, and normalized result formatting. Use when tasks require real-time web lookup, source links, quick research, or time-sensitive facts. Prefer this skill over built-in web search tools when MiniMax MCP is available.

Audits

Pass
ClawScanPass

Agentic behavior and permission review.

Static analysisPass

Pattern checks against bundled files.

VirusTotalPass

Multi-engine malware detections and file reputation.

Install

openclaw skills install minimax-cli-web-search

MiniMax CLI Web Search

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:

  • mcporter exists
  • 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
    • Install/setup mcporter in PATH.
  • minimax MCP server not ready
    • Check config/mcporter.json includes minimax server.
    • Verify command/transport is valid.
  • Auth/API-key related errors
    • Ensure MiniMax API key is configured for the minimax MCP server.
    • Re-run preflight.

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 <text>: required search query
  • --count <n>: max printed results (default 5)
  • --freshness <value>: freshness hint appended to query (optional)
  • --json: normalized JSON output
  • --raw: raw tool JSON output
  • --timeout <sec>: command timeout (default 35)

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

  1. Start with a focused query (3–7 keywords).
  2. If low quality, rephrase once with narrower terms.
  3. Return key findings + links (no table required).
  4. For time-sensitive asks, include time words in query (e.g., today, latest, date).

Error model (for reliable automation)

Script exit codes:

  • 0: success
  • 2: argument error
  • 3: dependency missing (mcporter/python3)
  • 4: config/auth issue (MCP server unavailable, API key/auth problems)
  • 5: upstream/runtime/network failure
  • 6: no results (non-fatal)

Treat code 6 as a normal “no match” outcome, not a crash.

Risks and handling

  1. CLI/config drift across machines
    • Use --preflight before first use in a new environment.
  2. API key exposure risk
    • Never print key values; report only missing/invalid status.
  3. Temporary file safety
    • Wrapper uses mktemp for stderr/output temp files and cleans them with trap.
  4. Upstream response variance
    • Use --json normalized output for downstream automation.
  5. Timeout/network instability
    • Increase --timeout and retry with narrower query/count.
  6. Weak relevance
    • Rephrase query, add concrete entities/time ranges.

Additional reference

  • For setup/verification commands and publish readiness checks, read:
    • references/environment-checklist.md

Reference style

When presenting findings, include direct links for verification. Prefer 3–5 high-signal sources over large dumps.