28 portable Agent Skills for structured reasoning in Claude Code, GitHub Copilot, Codex, Cursor, and other compatible tools
Claude Code Thinking Skills is a public catalog of Agent Skills. Use it for decisions, diagnosis, systems, risk, strategy, and innovation. Each skill gives an agent a clear procedure. Use a skill when its mechanism fits the task.
Use these skills when a problem needs a clear reasoning frame. The catalog covers diagnosis, decisions, creation, risk, execution, product, and strategy.
Read the catalog audit before making performance claims.
Claude Code can also load the same directories from ~/.claude/skills/ or .claude/skills/.
Use a skill
Start with the router when the right frame is unclear:
Use thinking-model-router to choose the right framework for this problem.
Invoke a leaf skill when the match is clear:
Use thinking-scientific-method to localize this bug.
Use thinking-reversibility to classify this architecture decision.
Use thinking-pre-mortem to stress-test this launch plan.
Use thinking-theory-of-constraints to find the binding bottleneck.
The Claude Code plugin uses thinking-skills:thinking-model-router as the exact router ID. The router can return NONE, one skill, or up to three complementary skills.
Skill catalog
The 28 active skill IDs match the directories under skills/
thinking-ooda; act while a time-pressured situation changes
Evidence
The catalog audit reports:
28 active skills, all manual-only
no automatic-retain verdict
portfolio-v1 with zero model calls and an unmeasured result
a provisional thinking-scientific-method row at +4.0 percentage points
The +4.0-point row is below the +5-point utility margin and has evidence gaps. Treat it as directional evidence, not an accuracy claim. See the decision-ready audit and analysis/evidence.json
Repository layout
skills/ Agent Skill source
scripts/ skill validation
evals/ structural, routing, and outcome evals
analysis/ evidence registry and audit
.claude-plugin/ Claude Code metadata
Study artifacts live under evals/studies/. External datasets follow their license limits.
Checks and contribution
Validate skill structure:
node scripts/validate-skills.js
Run the local structural gate:
EVAL_RUN=local node evals/run-structural.js
Run routing evaluation with the authenticated droid CLI:
To add a skill, create skills/thinking-{name}/SKILL.md with Agent Skills YAML front matter, a precise trigger, a non-trigger boundary, a procedure, and checks. Run validation and the relevant evaluation before submitting.
Questions
Which skill should I start with?
Use thinking-model-router when the fit is unclear. Use a leaf skill when the task already names its mechanism.
Can I use a skill without the router?
Yes. Invoke any skill directly when its trigger matches.
Do these skills guarantee better model accuracy?
No. The evidence supports structured procedures, not a guaranteed accuracy gain.
