Agent skills for Claude Desktop

code review feedback quality analysis best practices consistency checking

Use when writing or editing code comments in any language

Use when the user wants the LLM to do its own ML research: a fully-autonomous loop that hacks the training code, runs it, and keeps…

Use when writing, reviewing, or committing code to enforce Karpathy's 4 coding principles — surface assumptions before coding, keep it…

Use when writing, reviewing, or committing code to enforce Karpathy's 4 coding principles — surface assumptions before coding, keep it…

Evaluates code implementations against the minimalist and rigorous engineering principles championed by Andrej Karpathy.

Use this when: red-team my optimization metric, find ways to game my metric, metric pre-mortem, adversarial metric evaluation, gaming…

Build and maintain a persistent, compounding knowledge base as interlinked markdown files (Karpathy's LLM Wiki pattern): ingest sources…

Creating flashcards for law studies and exam preparation: Use case: Student wants definitions of legal concepts, norms, and…

Designs and generates minimalist 16:9 article illustrations featuring the signature Kasadin IP character for AI content and tutorials.

Achieve certain comprehension after AI work.

Comprehensive skill for the `kb` CLI and the Karpathy Knowledge Base pattern.

Use when creating, updating, or deleting documents in an sbdb-managed knowledge base.

Lint and health-check the knowledge base.

Ingest a source (URL, file, or freeform note) into the knowledge base.

Use ONLY for code review and test write/review guidance in `packages/agent-core-v2` (the DI × Scope agent engine).

Use when working with KDB-X Python for Python-to-q data conversion, querying tables with Column API, IPC connections, DB management, or…

Use when packaging code, datasets, configs, and deployment evidence for a KDD paper, where the repository cited in the submission is the…

Use when drafting KDD rebuttals on OpenReview, where authors answer each review under a no-hyperlink rule, cannot upload revisions, and…

Use when preparing an accepted KDD paper for ACM proceedings, covering the 12-page proceedings budget with 9 content pages, the extra…

Use when deciding whether a project belongs at KDD and in which track — Research vs Applied Data Science vs Datasets and Benchmarks vs AI…