hyperb1iss/hyperskills
14 skills4.2k installs350 starsGitHub
Install
npx skills add https://github.com/hyperb1iss/hyperskillsSkills in this repo
1Tui DesignTui-design is a Claude Code skill from hyperb1iss/hyperskills that functions as a design-pattern gallery for terminal user interfaces. It analyzes production TUIs such as lazygit—a Go gocui fork with five left panels plus a right detail column—and extracts layout, focus, and keybinding innovations like contextual footer actions that change per focused panel. Patterns cover persistent multi-panel layouts, popup layering, and zero-memorization interaction goals. Reach for tui-design when planning a new CLI dashboard, refactoring panel navigation, or benchmarking spatial organization against established tools before writing gocui, bubbletea, or similar terminal UI code.984installs2Orchestrateorchestrate is a hyperb1iss/hyperskills skill for multi-agent work at scale, mined from 597+ real agent dispatches across production codebases. It maps orchestration strategy to work shape, prompt structure to agent type, and background versus foreground execution to dependency graphs. Developers reach for orchestrate when running research swarms, parallel feature builds, wave-based dispatch, or build-review-fix pipelines that need fan-out coordination. Trigger phrases include swarm, parallel agents, multi-agent, orchestrate, fan-out, wave dispatch, and research army. The core principle is matching strategy to work through partitioning rather than single-agent serial execution.633installs3Researchresearch is a hyperb1iss/hyperskills agent workflow for large-scale knowledge gathering before decisions. The readme describes wave-based research with deferred synthesis, mined from 300+ real research dispatches where breadth-first gathering beat premature conclusions. Agents activate on investigate, evaluate options, state of the art, competitive analysis, codebase archaeology, or compare alternatives prompts. Core insight: research breadth-first, synthesize after—avoid locking conclusions from the first three sources. Developers reach for research when picking frameworks, evaluating LLM stacks, or mapping competitor capabilities and need structured multi-source intelligence. The pattern supports technology evaluation, SOTA analysis, and deep landscape exploration requiring multiple sources.633installs4Brainstormbrainstorm is a hyperb1iss/hyperskills workflow for structured ideation before creative work—new features, architecture decisions, project inception, or design exploration. The skill applies the Double Diamond model with persistent memory, mined from 100+ real brainstorming sessions across production projects. AI handles divergent phases for volume and cross-domain connections while humans own convergent judgment and selection. Developers reach for brainstorm when triggers like "let's think about," "how should we approach," or "explore options" appear instead of jumping straight into implementation.595installs5Planplan is a Structured Planning skill in hyperb1iss/hyperskills that decomposes complex work into verifiable tasks before any code is written. The workflow was mined from 200+ real planning sessions where plans survived contact with code only when every step had a concrete check rather than abstract bullets. Plans persist in Sibyl-native tracking so they outlive a single agent context window. Activate on phrases like write a plan, break this down, task decomposition, or spec this out. Reach for plan when a feature or epic is too large to implement in one agent session and you need a checklist the agent can verify step by step.593installs6Implementimplement is a Claude skill grounded in quantitative benchmarks from 21,321 tracked operations across 64+ projects in the Sibyl dataset. It teaches agents when to read versus search versus edit, how many verifications to run between commits, and how to avoid debugging spirals during feature work. Key metrics include 0.8 reads per code change, 0.5 searches per code change, a 2-3 edit sweet spot between verifications, 48.7 changes per commit, and 23 verifications per commit. Developers reach for implement when agent sessions feel chaotic—too many blind edits, missing test runs, or endless fix loops—and they want session discipline backed by real telemetry. The skill suits multi-step builds where verification cadence and exploration ratios materially affect merge quality.568installs7TiltCovers automating the local Kubernetes dev loop with Tilt (watch, build, deploy) via a Starlark Tiltfile. A developer uses it to write a Tiltfile, set up live_update, add resources, or debug a running Tilt instance.36installs8GitCovers advanced git workflows and conflict-resolution strategies. A developer uses it for rebases, merge conflicts, cherry-picks, bisect, or repository history work.30installs9DreamRuns a two-phase sleep cycle that extracts structured knowledge from past conversations and consolidates it into Sibyl. A developer uses it to capture decisions, patterns, and anti-patterns before sessions scroll off.17installs10TyCovers type checking Python code and setting up an LSP with ty. A developer uses it for ty check/server, resolving type errors, or configuring [tool.ty] in a project.17installs11UvCovers Python package and project management with uv via a workflow decision tree. A developer uses it for uv add/sync/run/lock/init, standalone scripts, workspaces, or managing Python versions.17installs12Uv BuildCovers building and publishing pure-Python packages with Astral's Rust-based uv_build backend. A developer uses it to configure a pyproject.toml build-system, produce sdist/wheel, or publish to PyPI.17installs13RuffCovers linting, formatting, and analyzing Python with ruff. A developer uses it for ruff check/format/fix, configuring ruff.toml, noqa/per-file-ignores, or the built-in ruff server.16installs14Cross Model ReviewRuns cross-model code review where the authoring model's code is reviewed by a different model with different failure modes. A developer uses it for an unbiased second opinion on code or a PR.14installs