turbo
Skill1 starsChanged 4 months agoMITView on GitHub
Generates code via Cerebras hosted LLM from a structured contract prompt, writes files directly to disk, then applies surgical Claude fixes for test failures (~60% token savings vs Claude direct). Use when you have a single task — algorithmic code, boilerplate, data transformations, or multi-file scaffolding — and want fast token-efficient generation.
- Reads credentials
4 other projects use a version of this.
Adopted by
Most recently taken up by
- smnandre/symfony-ux-skillstheir own version, first seen Nov 15, 2026
- oberemokis/skillstheir own version, first seen Nov 15, 2026
- tuanductran/hr-skillstheir own version, first seen Sep 9, 2026
- BKR-57/symfony-ux-skillstheir own version, first seen Mar 2, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- smnandre/symfony-ux-skills★ 186same name
- tuanductran/hr-skills★ 58same name
- oberemokis/skills★ 2same name
- BKR-57/symfony-ux-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- any-percentSkill
- judgeSkill
- showdownSkill
- speed-runSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 91
Run the test suite against generated code.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Turbo
Direct code generation via hosted LLM. Claude writes the contract, Cerebras implements the code, files are written directly to disk.
Announce: "I'm using speed-run:turbo for hosted code generation."
When to Use
Use turbo for:
- Algorithmic code (rate limiters, parsers, state machines)
- Multiple files (3+)
- Boilerplate-heavy implementations
- Token-constrained sessions
Use Claude direct instead for:
- CRUD/storage operations (Claude is cheaper due to no fix overhead)
- Single implementation with complex coordination
- Speed-critical tasks where fix cycles are costly
Tradeoffs
| Aspect | Claude Direct | Turbo (Hosted LLM) |
| Speed | ~10s | ~0.5s |
| Token Cost | Higher | ~60% savings |
| First-pass Quality | ~100% | 80-95% |
| Fixes Needed | 0 | 0-2 typical |
Workflow
Step 1: Write Contract Prompt
Structure your prompt with exact specifications:
Build [X] with [tech stack].
## DATA CONTRACT (use exactly these models):
[Pydantic models / interfaces with exact field names and types]
Example:
class Task(BaseModel):
id: str
title: str
completed: bool = False
created_at: datetime
class TaskCreate(BaseModel):
title: str
## API CONTRACT (use exactly these routes):
POST /tasks -> Task # Create task
GET /tasks -> list[Task] # List all tasks
GET /tasks/{id} -> Task # Get single task
DELETE /tasks/{id} -> dict # Delete task
POST /reset -> dict # Reset state (for testing)
## ALGORITHM:
1. [Step-by-step logic for the implementation]
2. [Include state management details]
3. [Include edge case handling]
## RULES:
- Use FastAPI with uvicorn
- Store data in [storage mechanism]
- Return 404 for missing resources
- POST /reset must clear all state and return {"status": "ok"}Step 2: Generate Code
mcp__speed-run__generate_and_write_files
prompt: [contract prompt]
output_dir: [target directory]Returns only metadata (files written, line counts). Claude never sees the generated code.
Step 3: Run Tests
Run the test suite against generated code.
Step 4: Fix (if needed)
For failures, use Claude Edit tool for surgical fixes (typically 1-4 lines each).
Common fixes:
| Error Type | Frequency | Fix Complexity |
| Missing utility functions | Occasional | 4 lines |
| Logic edge cases | Occasional | 1-2 lines |
| Import ordering | Rare | 1 line |
Step 5: Re-test
Repeat Steps 3-4 until all tests pass. Even with fixes, total token cost is much lower than Claude generating everything.
What Hosted LLM Gets Right (~90%)
- Data models match contract exactly
- Routes/endpoints correct
- Core algorithm logic
- Basic error handling
Configuration
| Variable | Default | Description |
CEREBRAS_API_KEY | (required) | Your API key |
CEREBRAS_MODEL | gpt-oss-120b | Model to use |
Available models:
| Model | Price (in/out) | Speed | Notes |
gpt-oss-120b | $0.35/$0.75 | 3000 t/s | Default - best value, clean output |
llama-3.3-70b | $0.85/$1.20 | 2100 t/s | Reliable fallback |
qwen-3-32b | $0.40/$0.80 | 2600 t/s | Has verbose <think> tags |
llama3.1-8b | $0.10/$0.10 | 2200 t/s | Cheapest, may need more fixes |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

