Engineering Skills · 技

Waza

Engineering habits you already know, turned into skills any AI agent can run.
Tw93
v3.28
MIT
8
engineering skills
6
agents supported
1
line to install
0
model lock-in

Models and tools come and go; the way good engineers actually work, planning, reviewing, debugging, reading the source, does not. Waza turns those habits into eight skills that run on whatever agent you have, so your workflow never depends on a single model staying available.

Plan & build

/think
pressure-test a task into a plan
/design
distinctive UI, never defaults
/learn
map a new domain fast
/read
distill any URL or PDF

Review & ship

/check
review the diff with evidence
/hunt
find a bug's true root cause
/write
make prose sound human
/health
audit your agent setup

Habits, not superpowers

The fashionable move is to bolt on a pack of "superpowers", dozens of skills promising an agent can suddenly do anything. But unconstrained capability is the problem, not the cure: it drifts into confident, generic, subtly wrong work. Waza adds no new powers. It encodes the habits a good engineer already has, so the model spends its capability on precision instead of breadth. Eight skills, not eighty, the few that actually decide quality, each carrying reference docs and real-failure gotchas rather than a one-line promise that rots the moment a task gets specific.

INPUT Engineering habits plan · review · debug · read WAZA · 8 SKILLS Goal + guardrails then let the model run OUTPUT Precise, reviewed work on any agent, not generic
Capability is the input, not the product. Constraints are what make it land.
Runs on Claude Code, Codex, Antigravity, OpenCode, Claude Desktop, and Pi. One workflow on every agent, installed once with npx skills add tw93/Waza.