Minimax Web SearchSAFE
High-performance AI coding agent CLI written in Rust with zero unsafe code
Overview
High-performance AI coding agent CLI written in Rust with zero unsafe code
46a2537504c9OBSERVED · 2026-10-10Host compatibility
What the documentation claims. We have not run a compatibility test.
| Host | Status | Notes |
|---|---|---|
| claude-code | mentioned |
What it tells the agent
The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.
---
name: minimax-web-search
description: Search the web for current information using the web_search tool
---
# MiniMax Web Search Skill
Use this skill when you need to find current, real-time, or authoritative information that may have changed since my knowledge cutoff.
## How to Use
Call the `web_search` tool directly with a query:
```
web_search({ query: "your search query" })
```
## When to Use
Use `web_search` when:
- **Current events**: News, recent developments, latest releases
- **Fact verification**: Checking if information is still accurate
- **Technical updates**: New versions, patches, breaking changes
- **Dynamic content**: Prices, availability, schedules
- **Authoritative sources**: Official documentation, API references
## When NOT to Use
Do NOT use `web_search` when:
- **General knowledge** I likely know (historical facts, basic concepts)
- **Coding help** for common patterns (unless asking about latest practices)
- **Opinion-based questions** without specific recent information needs
- **Files already in context** or accessible via `read` tool
## Usage
```
web_search({
query: "your search query here"
})
```
## API Details
**Endpoint**: `POST {api_host}/v1/coding_plan/search`
**Request Body**:
```json
{
"q": "your search query"
}
```
**Response Format**:
```json
{
"organic": [
{
"title": "Result Title",
"link": "https://example.com",
"snippet": "Brief description...",
"date": "2025-01-15"
}
],
"related_searches": [
{"query": "related query 1"},
{"query": "related query 2"}
],
"base_resp": {
"status_code": 0,
"status_msg": "success"
}
}
```
## Tips for Better Results
1. **Be specific**: "TypeScript 5.4 new features" vs "TypeScript updates"
2. **Include context**: "React 19 server components vs client components"
3. **Use recent keywords**: "2024" or "2025" when timing matters
4. **Target authoritative sources**: Include product names, official sites
5. **Use 3-5 keywords** for optimal results
## Examples
### Good Queries
- "Node.js 22 new features"
- "Claude Code vs pi coding agent 2025"
- "React 19 release date official"
- "Docker desktop alternatives 2024"
- "TypeScript 5.4 generics tutorial"
### Poor Queries
- "how to use TypeScript" (too general)
- "is Python popular" (opinion, not time-sensitive)
- "help me debug my code" (not a search task)
- "best programming language" (opinion-based)
## Response Format
Results include:
- **Titles** and **URLs** for each result
- **Snippets** with relevant context
- **Dates** when available
- **Related searches** for follow-up queries
You can use the suggestions to refine your search or ask the user which direction they'd like to explore.
## Error Handling
- **Status code 1004**: Authentication error - check API key and region
- **Status code 2038**: Real-name verification required
- **Other errors**: Check Trace-Id in error message for supportTrust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (1)
.beads/beads.jsonl
Gates applied: no_behavioural_pass.
46a2537504c9full audit observations/trust-audit/skill/dicklesworthstone__minimax-web-search.json · Report an issue / request a re-scanAudit history
Every audit this skill has had.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-10 | 46a2537504c9 | SAFE | B | 89 | first audit |
Questions
What does the Minimax Web Search skill do?
High-performance AI coding agent CLI written in Rust with zero unsafe code
Is Minimax Web Search safe to install?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean skill reads B.
What can Minimax Web Search access on my machine?
The audit observed no filesystem, network or shell use at all in its source.
Which assistants does Minimax Web Search work with?
Its documentation mentions claude-code. That is what the text claims, not a compatibility test we ran.
How current is this page?
The grade is for one exact copy of the source (46a2537504c9), read on 2026-10-10. The repository is watched, and a new audit runs when it changes — this is the first audit.