| name | sauce |
| version | 1.0.0 |
| description | Find best-in-class tech for any problem. Runs parallel searches across GitHub, startups, research, and community — ranks and saves winners to a sauce list.
|
| allowed-tools | ["Bash","Read","Write","Edit","Grep","Glob","Agent"] |
/sauce — Find the Best Tech
Given a problem, run broad parallel searches and surface the best options. Save winners to a persistent list.
Commands
| Input | Action |
|---|
/sauce <problem> | Research mode |
/sauce list | Show saved sauce list |
/sauce save <name> | Save a find to the list |
/sauce remove <name> | Remove from list |
Research mode
Step 1: Understand the problem
- Core requirement
- Key constraints (performance, cost, simplicity, ecosystem)
- What they've tried
Step 2: Generate 15–20 diverse queries
Cover these angles:
- Direct — exact problem statement, year (
2025 2026)
- GitHub —
github stars open source alternative
- Startups —
YC backed series A startup
- Research —
benchmark comparison paper
- Community —
HN discussion production experience
Step 3: Run searches in parallel
Spawn 3–4 agents concurrently, each handling a query category. Use Exa MCP tools:
- General + comparisons →
mcp__claude_ai_Exa__web_search_exa
- Companies + startups → same with
category: "company"
- Deep research →
mcp__claude_ai_Exa__web_search_exa with type: "deep"
- Code/GitHub →
mcp__claude_ai_Exa__web_fetch_exa for specific repos
Agent prompt template:
You are a tech research agent. Search for solutions to: "<PROBLEM>"
Run these queries using mcp__claude_ai_Exa__web_search_exa (numResults: 8, type: "auto"):
1. "<query1>"
2. "<query2>"
...
For each notable find, return: name, URL, one-line description, why it's exceptional, traction signals (stars, funding, usage), maturity (alpha/stable/production).
Step 4: Score and rank
For each unique technology:
| Dimension | Weight |
|---|
| Relevance — actually solves the problem | 3× |
| Maturity — production-ready, actively maintained | 2× |
| Traction — stars, funding, adoption | 2× |
| Innovation — genuinely 10× better approach | 2× |
| Ecosystem — docs, integrations, community | 1× |
Composite = weighted sum / 10.
Step 5: Present
## Sauce: "<problem>"
Searched 18 queries. Found 12 unique technologies.
### 1. Technology Name (9.2/10)
**What:** one-line description
**Why it's sauce:** the 10× factor
**URL:** https://...
**Traction:** 20k stars, backed by CNCF, used in production at Cloudflare
**Maturity:** stable
### 2. ...
---
### Honorable Mentions
- **Tech A** — interesting but alpha (URL)
Save anything? Run `/sauce save <name>`.
Sauce list
Lives at skills/sauce/references/sauce-list.md.
List mode
Read and display the file grouped by status (adopted / exploring / rejected).
Save mode
Append to sauce-list.md:
## Technology Name
- **Solves:** problem description
- **Why:** what makes it exceptional
- **URL:** primary link
- **Discovered:** YYYY-MM-DD
- **Status:** exploring
- **Notes:** context
Rules
- Breadth first — diverse queries before narrowing
- Prefer technologies active in the last 12 months
- Traction matters — 10 GitHub stars is not sauce unless brand new and exceptional
- No vaporware — must have working code, live product, or published paper
- Be opinionated — the top pick should be obvious
- Deduplicate — same tech found multiple ways counts once