andrej-karpathy-skills
- Repo stars 83
- Author updated Live
- Author repo oh-my-kimi
- Domain
- AI
- Compatible agents
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- Trust score
- 88 / 100 · community maintained
- Author / version / license
- @dmae97 · no license declared
- Token usage
- Lean
- Setup complexity
- Plug-and-play
- External API key
- Not required
- Operating systems
- Unspecified (assume cross-platform)
- Runtime requirements
- No special requirements
- Permissions
-
- Read-only
- Write / modify
- Network behavior
- Local-only
- Install commands
- 26 variants
Profile is derived at build time from SKILL.md and install vectors. Subject to drift from author intent.
Heads up: 未限定 allowed-tools，默认拥有全部工具权限。
---
name: andrej-karpathy-skills
description: Minimal, goal-driven, surgical coding workflow adapted from forrestchang/andrej-karpathy-skills…
category: ai
runtime: no special runtime
---
# andrej-karpathy-skills output preview
## PART A: Task fit
- Use case: Minimal, goal-driven, surgical coding workflow adapted from forrestchang/andrej-karpathy-skills for OMK. Use for coding, refactoring, debugging, and review tasks where assumptions, overengineering, or broad edits could cause regressions..
- Inputs: target material, constraints, expected output, and acceptance criteria.
- Evidence boundary: follow “Use when / OMK workflow / Output contract” and do not present inference as author intent.
## PART B: Execution result
- **01** The card summarizes the use case; runtime output centers on “Minimal, goal-driven, surgical coding workflow adapted from forrestchang/andrej-karpathy-skills for OMK. Use for coding, refactoring, debugging, and review tasks where assumptions, overengineering, or broad edits could cause regressions.”.
- **02** When the source has headings, the agent prioritizes “Use when / OMK workflow / Output contract” so the result follows the author’s structure.
- **03** Typical output includes task judgment, concrete steps, required commands or file edits, validation, and follow-up options.
- **04** Risk context follows the fingerprint: read files, write/modify files; mostly runs locally; usually needs no extra API key.
## Running Rules
- read files, write/modify files; mostly runs locally; usually needs no extra API key.
- Validate with a small sample before expanding scope.
- Return the result, validation criteria, and next iteration options. Choose the trigger
The source does not require a stable slash command. After installation, invoke the skill by name and describe the task.
Provide inputs and boundaries
Name target files or source material, expected output, forbidden changes, and whether network or shell access is allowed. Permission fingerprint: read files, write/modify files.
Validate with a small run
Start with a small task and check whether the result follows “Use when / OMK workflow / Output contract”. Inspect diffs, logs, previews, or tests before expanding scope.
Review before delivery
Confirm the final output includes a concrete result, evidence, and next action. If it stays generic, tighten inputs, boundaries, and acceptance criteria.
---
name: andrej-karpathy-skills
description: Minimal, goal-driven, surgical coding workflow adapted from forrestchang/andrej-karpathy-skills…
category: ai
source: dmae97/oh-my-kimi
---
# andrej-karpathy-skills
## When to use
- Minimal, goal-driven, surgical coding workflow adapted from forrestchang/andrej-karpathy-skills for OMK. Use for codin…
- Use it when the task has clear inputs, repeatable steps, and validation criteria.
## What to provide
- Target material, scope, expected result, and forbidden changes.
- Whether network, commands, file writes, or external services are allowed.
## Execution rules
- Organize steps around “Use when / OMK workflow / Output contract” and keep inference separate from source facts.
- read files, write/modify files; mostly runs locally; usually needs no extra API key.
- Validate with a small sample before expanding the task.
## Output requirements
- Return the deliverable, key evidence, validation method, and next action.
- Mark missing information as unknown; do not invent commands, platforms, or dependencies. Evidence boundary and execution chain
The author source anchors workflow facts; repository files anchor sources and commands; Fluxly only adds fit, limitations, and quality judgment.
skill "andrej-karpathy-skills" {
input -> user goal + target files + boundaries + acceptance criteria
context -> Use when / OMK workflow / Output contract
rules -> SKILL.md triggers / order / output contract
runtime -> no special runtime | read files, write/modify files | mostly runs locally
guardrails -> usually needs no extra API key + small-sample validation + diff/log review
output -> copyable result + checklist + next iteration
}andrej-karpathy-skills
Source basis: forrestchang/andrej-karpathy-skills at commit 2c606141936f1eeef17fa3043a72095b4765b9c2. This OMK skill is a compact adaptation, not a vendored copy of upstream prompts or code.
Use when
- Any non-trivial code change needs tight scope and proof.
- The request is ambiguous enough that silent assumptions could be wrong.
- A refactor or bugfix risks drive-by edits.
- You need to turn an imperative request into verifiable success criteria.
OMK workflow
- State assumptions and tradeoffs only when they affect the implementation path. Ask if ambiguity blocks safe progress.
- Define success as observable checks: tests, typecheck, lint, build, screenshots, replay, or exact output.
- Make the smallest change that satisfies the goal. Avoid speculative abstractions and features not requested.
- Touch only files directly tied to the request. Preserve existing style and avoid unrelated cleanup.
- Remove only dead code introduced by your own change unless the user asked for broader cleanup.
- Verify, inspect the diff, and report remaining risks honestly.
Output contract
Return:
- assumptions that mattered
- success criteria and checks
- changed files
- commands run
- pass/fail status
- risks or blocked items
Guardrails
- Do not broaden scope to improve adjacent code.
- Do not hide confusion behind a confident implementation.
- Do not claim completion without evidence.
Decide Fit First
Design Intent
How To Use It
Boundaries And Review