Skills for AI coding assistants. The core is a tightly-coupled workflow chain for structured development (applicable to any task requiring analysis, planning, and systematic execution); standalone skills handle on-demand work like documentation governance.
- Concise: Each skill is focused and minimal — no heavy checklists or verbose gates
- Self-driving chain: Workflow skills form an autonomous chain; each drives to the next without an orchestrator
- Manual control: Skills are user-triggered, not auto-injected via hooks
The collection spans two categories: workflow skills that form a self-driving chain, and standalone skills invoked on demand.
| Skill | Description |
|---|---|
using-devflow |
Entry router — skill index, workflow orchestration, intent routing |
brainstorming |
Requirements → design → long-lived docs (Spec/PRD/ADR) |
planning |
Design doc → milestone-grouped task list + execution strategy (TDD decisions) |
executing-plans |
Serial milestone execution with checkpoint reviews |
test-driven-development |
Red-Green-Refactor cycle, invoked on-demand by executing-plans |
verifying-completion |
Run commands → see output → then claim completion |
reviewing-code |
Dispatch reviewer → structured report + handle feedback |
finishing-work |
Integration decision: merge, PR, keep, or discard |
systematic-debugging |
Root cause investigation → hypothesis → minimal fix |
| Skill | Description |
|---|---|
curating-docs |
Tame documentation chaos — survey, recommend convention, codify into AI config, reorganize structure, remediate content health |
flowchart LR
A[using-devflow] --> B[brainstorming]
B --> C[planning]
C --> D[executing-plans]
D --> E[finishing-work]
D -.->|per task| F[verifying-completion]
D -.->|per milestone + final| G[reviewing-code]
using-devflow auto-drives the chain without per-step confirmations. systematic-debugging and reviewing-code can also be invoked independently.
Inspired by Superpowers.