Skills
Attachable specialists and modifiers that directly shape runtime work.
See the skill catalogAgent Skills library
Attach the right specialist to the task.
Amplifiers packages specialist instructions, deep references, and optional helpers the way modern skill runtimes expect: one skill per folder, one clear runtime entrypoint, and deeper material loaded only when needed. The model follows the Agent Skills pattern used across Anthropic, OpenAI Codex, and related tooling.
- Runtime skills live in
skills/ - Recurring combos live in
stacks/ - Deep material lives in each skill's
references/ - Public docs site lives in
docs/
Repository model
One canonical skill layer. Everything else supports it.
Stacks
Reusable combinations for delivery types that repeatedly need multiple specialists.
See the stack catalogSkill packages
Each skill keeps its own references, scripts, assets, and optional runtime metadata.
See the package anatomyHow to work with it
Start from the delivery, then attach the specialists you need.
1
Choose a single skill when the task is narrow and needs one clear specialist.
2
Attach multiple skills when the delivery spans design, writing, and code.
3
Reuse a stack when the same combination keeps showing up in daily work.