Software Development Skills

Use to launch, monitor, and manage long-running background tasks — training, evaluation, batch processing. Provides nohup + log + PID file pattern, tmux integration, monitoring helpers. Triggers запусти в фоне, run in background, мониторь, что там, kill процесс.

Build a custom browser-based annotation interface tailored to your data for reviewing LLM traces and collecting structured feedback. Use when you need to build an annotation tool, review traces, or collect human labels.

Analyze a repository to generate recommended Claude Code settings.json permissions. Use when setting up a new project, auditing existing settings, or determining which read-only bash commands to allow. Detects tech stack, build tools, and monorepo structure.

Systematic techniques for reading and understanding unfamiliar legacy code without…

Create a git branch following Sentry naming conventions. Use when asked to "create a branch", "new branch", "start a branch", "make a branch", "switch to a new branch", or when starting new work on the default branch.

Admin-role delivery skill for framework/docs repos. Executes the full engineer delivery flow (simplify → quality → behavioral verify → review → rebase → commit → PR) without JIRA ceremony. Trigger: '發 PR', 'open PR', 'create PR', 'PR workflow', '準備發 PR'. For product repos with JIRA tickets, use engineering instead. PR body logic is in references/pr-body-builder.md (shared by both roles).

Build Gradio web UIs and demos in Python. Use when creating or editing Gradio apps, components, event listeners, layouts, or chatbots.

Use Hugging Face Hub via MCP server tools. Search models, datasets, Spaces, papers. Get repo details, fetch documentation, run compute jobs, and use Gradio Spaces as AI tools. Available when connected to the HF MCP server.

Use when looking for best HuggingFace models, datasets, or spaces by criteria. Triggers: best model, top model, recommend hf model. for a task, wants to know what AI model to use, or wants to compare models by benchmark scores. Triggers on: "best model for X", "what model should I use for", "top models for [task]", "which model runs on my laptop/machine/device", "recommend a model for", "what LLM should I use for", "compare models for", "what''''s state of the art for", or any question about choosing an AI model for a specific use case. Always use this skill when the user wants model recommendations or comparisons, even if they don''''t explicitly mention HuggingFace or benchmarks.

Run evaluations for Hugging Face Hub models using inspect-ai and lighteval on local hardware.

Use when working with HuggingFace datasets library — loading, processing, uploading. Triggers: loaddataset, hf datasets, dataset upload. subset/split metadata, paginate rows, search text, apply filters, download parquet URLs, and read size or statistics.

Publish and manage research papers on Hugging Face Hub.

Use when building tools/scripts that combine multiple HuggingFace API calls. Triggers: hf tool, hf api workflow, hub api scripts. task where using data from the Hugging Face API would help. This is especially useful when chaining or combining API calls or the task will be repeated/automated. This Skill creates a reusable script to fetch, enrich or process data.

Track and visualize ML training experiments with Trackio.

Iterate on a PR until CI passes. Use when you need to fix CI failures, address review feedback, or continuously push fixes until all checks are green.

Use when writing GitHub pull request description following conventions. Triggers: write PR, PR description, open pull request. opening a PR or refreshing an existing PR after material changes.

Use before any git commit or PR. Runs format/lint/typecheck/tests, checks for secrets/large files/debug statements, validates commit message. Triggers готов к commit, проверь перед commit, можно ли пушить, ready to merge.

Three orthogonal analytical operations (WHERE/WHEN/WHY) + cross-operation synthesis. Each operation attacks the problem from a fundamentally different angle. The disagreements between the three ARE the valuable output. Works on any domain — code, business, strategy, design, text.

Discover all possible analysis domains for an artifact. Finds obvious and non-obvious angles — architecture, security, but also marketing positioning, user psychology, regulatory implications, teaching value. Use before prism-scan or prism-full to explore what's worth investigating.

Full Prism: multi-pass structural analysis with mandatory adversarial self-correction. Designs custom analytical passes, executes them with chaining, then attacks its own findings before synthesizing. Use for maximum depth on important code or artifacts.

Constraint transparency: analyzes an artifact structurally, then analyzes what its own analysis concealed. Produces a conservation law AND a constraint report showing what was maximized, what was sacrificed, and what to investigate next. The only AI skill that knows what it can't see.

Structural analysis through dynamically generated cognitive lenses. Generates the optimal analytical lens for the specific code/artifact, then executes it. Finds conservation laws, structural invariants, and concrete bugs that vanilla analysis misses. Use on any code file, system design, or text artifact.

Use AFTER successful fix to save it for next time. Triggers запомни как фиксили, сохрани рецепт, this fix worked, save recipe, добавь в память, save to memory, remember this fix. Кладёт SKILL.md-стиль recipe в .claude/memory/recipes/ + обновляет MEMORY.md.

Use BEFORE writing API keys/tokens/passwords into code, BEFORE git commit/push, BEFORE creating PR. Blocks accidental leaks of secrets to GitHub. Triggers коммит, push, ключ, токен, secret, api key, .env, credentials, push to github, открой PR, открыть PR, git commit, git push.