hyperskills
Focused AI agent skills for things models don't already know
What This Is
Models already know how to write React components, Kubernetes manifests, and PyTorch code. They don't need 300 lines of examples for that.
hyperskills provides skills for things that are genuinely hard to get right without guidance — procedural knowledge distilled from thousands of real sessions, decision trees for high-stakes operations, and multi-agent orchestration patterns that actually work in production.
10 skills. Zero bloat. Each one earned its place through real-world evidence.
Installation
Claude Code
/plugin install hyperskills
Vercel Skills (skills.sh)
# Install all skills
npx skills add hyperbliss/hyperskills --all
# Or pick what you need
npx skills add hyperbliss/hyperskills --skill implement
npx skills add hyperbliss/hyperskills --skill orchestrate
Manual
git clone https://github.com/hyperb1iss/hyperskills.git
ln -s $(pwd)/hyperskills/skills ~/.claude/skills/hyperskills
How the Skills Work Together
The skills form a workflow pipeline. Each one handles a phase of the development lifecycle and hands off to the next:
brainstorm ──→ research ──→ plan ──→ implement ──→ codex-review
│ │ │ │
│ │ │ └──→ git
│ │ │
│ │ └──→ orchestrate (parallel agents)
│ │
└──────────────┴──→ Any skill can loop back when new questions emerge
Typical flows:
| Scenario | Flow |
|---|---|
| New feature | brainstorm → plan → implement → codex-review |
| Greenfield project | brainstorm → research → plan → orchestrate → implement |
| Bug fix | implement (straight to it — scale selection handles this) |
| Architecture decision | brainstorm → research → decide |
| Large refactor | plan → orchestrate → implement → codex-review |
You don't need to follow the full pipeline. Each skill has built-in scale selection — a typo fix doesn't need brainstorming, and a clear bug doesn't need research. Start wherever makes sense.
Skills
Process Skills
These encode how to approach a class of work — workflows, phases, and decision gates.
brainstorm — Structured Ideation
Double Diamond model for creative work: diverge on the problem, converge on a definition, diverge on solutions, converge on a decision. Grounded in persistent memory (Sibyl) so you never re-explore solved problems. Includes a Council pattern for complex architectural decisions using advocate/critic agents.
/hyperskills:brainstorm
research — Multi-Agent Knowledge Gathering
Wave-based research with deferred synthesis. Deploy agents in waves across a research surface, run gap analysis between waves, then synthesize with the full picture. Covers technology evaluation, codebase archaeology, SOTA analysis, and competitive landscape patterns. Max 3 waves — if that's not enough, reframe the question.
/hyperskills:research
plan — Task Decomposition
Verification-driven planning where every step has a concrete check. Decomposes work into 2-5 minute tasks ordered by dependency, marks parallelizable waves for orchestration, and tracks everything in Sibyl. Includes a trust gradient — full ceremony for early tasks, lighter review as patterns stabilize.
/hyperskills:plan
implement — Verification-Driven Coding
The core implementation skill, distilled from 21,321 tracked operations across 64+ projects. Encodes the patterns that consistently ship working code: