| name | algo |
| description | Algorithm analysis with parallel multi-model reasoning — deepseek_algo (lead), qwen_algo, qwen_reason, and minimax_code tackle your problem from different angles |
| user-invocable | true |
Algo — Multi-Model Algorithm Analysis
Three specialized models attack your algorithm problem in parallel, then synthesize the best approach.
Usage
/algo [problem or code]
/algo [focus] [problem or code]
Focus: optimize, complexity, data-structure, memory, correctness, competitive, cache, general (default)
When to Use
- Algorithm design or optimization
- Competitive programming problems
- Complexity analysis (time/space)
- Data structure selection
- Code performance profiling
- "Is there a better approach?"
Instructions
When user invokes /algo [problem]:
Step 1: Parse Input
Extract:
- Focus — if specified (e.g.
/algo optimize [code]), use it. Default: general
- Problem — the algorithm question or code
- Constraints — if user mentions N size, time limits, memory limits, extract them
Step 2: Parallel Analysis (4 models)
Call ALL FOUR in parallel:
deepseek_algo — Lead algorithmic reviewer (DeepSeek V4 Pro, top AIME/CodeElo):
mcp__tachibot-mcp__deepseek_algo({
problem: "[problem/code]",
focus: "[focus]",
constraints: "[constraints if any]",
context: "[any additional context]"
})
qwen_algo — Algorithm specialist (complexity, data structures, competitive patterns):
mcp__tachibot-mcp__qwen_algo({
problem: "[problem/code]",
focus: "[focus]",
constraints: "[constraints if any]",
context: "[any additional context]"
})
qwen_reason — Mathematical reasoning (proofs, correctness, edge cases):
mcp__tachibot-mcp__qwen_reason({
problem: "Analyze this algorithm problem mathematically:\n\n[problem/code]\n\nFocus on:\n1. Correctness proof or counterexample\n2. Edge cases that break the solution\n3. Tight bound analysis (best/worst/average)\n4. Mathematical invariants",
approach: "mathematical"
})
minimax_code — Implementation review (MiniMax M3, 1M ctx, cheap second opinion):
mcp__tachibot-mcp__minimax_code({
query: "[problem description + focus]",
task: "optimize",
code: "[code if provided]",
language: "[language if detectable]"
})
If any tool is unavailable, proceed with the remaining ones. Minimum: any 1 model.
Step 3: Synthesize
Compare all three responses. Look for:
- Agreement — all 3 suggest the same approach? High confidence.
- Conflict — different approaches? Compare complexity bounds, pick the one with tighter guarantees.
- Unique insights — did one model catch something the others missed?
Step 4: Present Results
Format:
## Algorithm Analysis: [problem summary]
### Recommended Approach
**Algorithm:** [name/type]
**Time:** O([complexity])
**Space:** O([complexity])
**Why:** [1-2 sentences]
### Analysis Breakdown
#### Algorithmic Analysis (Qwen Algo)
[Key findings — complexity, data structure choice, optimization tiers]
#### Mathematical Reasoning (Qwen Reason)
[Correctness, edge cases, tight bounds, invariants]
#### Implementation (MiniMax)
[Code suggestions, practical optimizations, SWE perspective]
### Consensus
[What all models agreed on]
### Key Insight
[The most valuable finding — often from just one model]
### Edge Cases to Watch
1. [edge case 1]
2. [edge case 2]
### Alternative Approaches
| Approach | Time | Space | Trade-off |
|----------|------|-------|-----------|
| [Current] | O(...) | O(...) | [balanced] |
| [Alternative 1] | O(...) | O(...) | [faster but more memory] |
| [Alternative 2] | O(...) | O(...) | [simpler but slower] |
Examples
/algo implement LRU cache with O(1) operations
/algo optimize this sorting function: [paste code]
/algo competitive find longest palindromic subsequence, N<=5000
/algo complexity what's the tight bound for this recursive function
/algo data-structure best structure for range queries with updates