Individual Package Skills
| gsDesign |
gsDesign |
Classical group sequential boundaries, spending functions, survival trial design |
| gsDesign2 |
gsDesign2 |
Next-generation GSD with average hazard ratio, non-proportional hazards, spending time |
| graphicalMCP |
graphicalMCP |
Graphical multiple comparison procedures, Bonferroni-based closed testing |
| rpact |
rpact |
Confirmatory adaptive designs, sample size reassessment, combination tests |
| simtrial |
simtrial |
Survival trial simulation, weighted logrank, MaxCombo tests |
| wpgsd |
wpgsd |
Weighted parametric GSD with correlated test statistics |
| gsDesignNB |
gsDesignNB |
Negative binomial / recurrent event trial design and simulation |
| gMCPLite |
gMCPLite |
Legacy graphical MCP (prefer graphicalMCP for new projects) |
Cross-Package Skills
| graphicalMCP-gsDesign2 |
graphicalMCP + gsDesign2 |
GSD with graphical multiplicity control, Maurer-Bretz framework |
| illness-death |
gMCPLite (simulation code) |
Multi-state model for correlated OS/PFS/ORR endpoints |
| multi-endpoint-sim |
gsDesign + gMCPLite + simtrial + graphicalMCP |
Full simulation pipeline from design through multiplicity-adjusted testing |
Semantic Layer
| Semantic Layer |
all packages |
Model-neutral glossary, routing skill, gsDesignNB seed crosswalk, and routing examples |
Skill Anatomy
Each skill follows the same structure:
.agents/skills/<skill-name>/
├── SKILL.md # Triggers, workflow, key concepts
├── agents/
│ └── openai.yaml # Codex-compatible display metadata
└── references/
├── code_patterns.md # Verified code templates
└── llms.txt # API documentation (where available)
- SKILL.md defines when to activate and what domain knowledge to apply
- agents/openai.yaml provides display metadata for Codex-compatible skill surfaces
- Code patterns provide tested, working templates as starting points
- llms.txt files provide vendored API details where available
- .cursor/rules/gsdesign-skills.mdc adapts the portable skills for Cursor