You do not need another generic “be thoughtful” prompt. You need thinking skills the agent can run: brainstorm without shutting ideas down too early, stress-test a plan, map system effects, or flag harm before you ship.
Most chats mix those jobs in one reply. Ideas, worries, and a verdict show up together. Fine for a quick take, not if you want the same structure next week.
What this library is
AI Agent Skills packages ten frameworks as installable SKILL.md workflows for Cursor, Claude Code, and other tools that follow the Agent Skills standard. Each skill is a fixed phase sequence, not a vibe. Install once; when the task fits, the agent loads that pass instead of inventing a new outline every time.
New to skills? Start with Your AI Agent Shouldn't Start From Zero Every Session. Below: what each skill does, when to pick it, and how to install from GitHub or ClawHub.
From brainstorming to critical analysis
Real work jumps between modes:
- Brainstorm: creative and lateral thinking when you need volume or a pattern break.
- Frame and decide: design thinking for user problems, strategic thinking for bets, Six Thinking Hats when one decision needs facts, feelings, risks, upsides, and ideas kept apart.
- Analyze: analytical thinking for hypotheses and evidence; systems thinking for loops and side effects.
- Critique: critical thinking to audit one claim; ethical thinking when values and harm matter, not just legality.
- Rebuild: first principles when “how everyone does it” might be wrong.
One mega-prompt will not hold that stack. Ten skills let you call the mode you need by name or in plain language.
npx skills add ysskrishna/ai-agent-skills # all ten thinking skills
npx skills add ysskrishna/ai-agent-skills --skill creative-thinking # one skillAfter install, use slash commands like /creative-thinking where your host supports them, or say “use critical thinking” in the prompt.
Claude Code and gh skill install are documented in the repo README. OpenClaw users can browse ClawHub.
The ten thinking skills at a glance
| Skill | Reach for it when… | Default pass |
|---|---|---|
| six-thinking-hats | One decision needs multiple angles in order (facts, gut, risks, upsides, ideas) | Hat sequence (mode-dependent) |
| critical-thinking | Audit one claim, plan, or belief | Clarify → Information → Assumptions → Reasoning → Alternatives → Conclusion |
| systems-thinking | Feedback loops, delays, leverage, unintended consequences | Boundary → Structure → Dynamics → Delays → Leverage → Synthesis |
| creative-thinking | Brainstorm with volume before you commit | Prime → Diverge → Connect → Harvest |
| strategic-thinking | Bets under constraints (where to play, how to win) | Intent → Landscape → Advantage → Options → Choice → Risks & Cadence |
| analytical-thinking | Hypotheses, metrics, structured breakdown | Frame → Decompose → Hypotheses → Evidence → Synthesis |
| lateral-thinking | Ideation is stuck or incremental | Po → Extract principle → Concept fan → Candidates |
| design-thinking | Human-centered discovery before build | Empathize → Define → Ideate → Prototype → Test plan |
| first-principles-thinking | Strip convention and rebuild from bedrock | Surface → Question → Bedrock → Rebuild → Implications |
| ethical-thinking | Values, harms, consent, power (beyond legality) | Stakeholders → Values → Harms/Benefits → Justice/Power → Options → Recommendation |
Which skill when?
| If you need… | Start here |
|---|---|
| Multiple angles on one decision, without one muddy paragraph | six-thinking-hats |
| A verdict on one proposal (evidence, assumptions, logic) | critical-thinking |
| Many ideas before narrowing | creative-thinking |
| To break a rigid pattern | lateral-thinking |
| User problem clarity before solutions | design-thinking |
| Roadmap or portfolio bets | strategic-thinking |
| Metrics and hypotheses | analytical-thinking |
| Second-order effects | systems-thinking |
| To challenge industry defaults | first-principles-thinking |
| Stakeholder harm and fairness | ethical-thinking |
Trust before you install
Read each SKILL.md, skim any scripts, try in a throwaway project. These packs are instruction-only; the habit still matters.
What to do next
- Pick a row in “which skill when?” that matches something on your desk.
npx skills add ysskrishna/ai-agent-skills --skill <name>and describe the task in plain language.- For hat-by-hat decisions, read Six Thinking Hats.
