Agent Skills by Sujin Hwang
Showing 18 vetted skills indexed across 1 GitHub repositories.
skill-transplant
Move a feature and its dependencies between repositories using a five-phase protocol.
map-workflow
Map a domain's workflow into a load-ready domain-workflow.md document.
design-backtesting
Define backtesting library artifacts and sequential rules for OS calibration.
design-loops
Design compounding feedback loops that accumulate operational data for agent OS performance.
design-identity
Design a user identity model with axes, personas, and privacy rules for the target OS.
design-localization
Build a localization framework for multi-language operating systems.
discover-domain
Conduct an interactive domain discovery interview to populate domain-input files.
design-data-sources
Map external data sources with trust levels, freshness thresholds, and fallback chains.
generate-os
Assemble domain inputs and designs into a deployable Claude Code OS repository.
design-skills
Design a documented v1 skill set for an agent OS.
design-sub-agents
Design adversarial reviewer personas that critique creation skill outputs for edge cases.
design-artifact-pipeline
Design artifact pipelines to standardize generation, rendering, and verification of non-text deliverables.
design-data-contract
Inventory and classify agent OS files by ownership type into a data contract.
write-quality-contract
Draft a quality-contract specification codifying rules, checks, anti-patterns, and tone for an agent OS.
define-audiences
Identifies stakeholders and documents evaluation criteria for reviewer sub-agent design.
design-batch-architecture
Designs optimal batch workflows for parallel item processing with deterministic state tracking and merge semantics.
build-scoring
Design multi-dimensional evaluation rubrics with calibrated scales, thresholds, and adaptive weights.
review-os
Audit generated OS repositories against 15 design principles and a 23-item checklist.