z is a five-role AI delivery plugin built from a mixed operating model:
- Light orchestration from startup-friendly agent teams.
- Heavy review, QA, and recovery gates from verification-first teams.
The result is a plugin that stays small enough for founder-led execution while still enforcing independent review, diagnosis-before-fix, and reusable knowledge capture.
z ships with five core roles:
@z-lead- intake, scope control, routing, handoffs.@z-planner- clarification, task breakdown, acceptance criteria, risk.@z-implementer- scoped code changes and local verification.@z-reviewer- requirement, security, and out-of-scope checks.@z-qa-debugger- testing, repro, diagnosis, and re-verification.
z uses a compact six-step workflow:
- Intake: capture objective, constraints, owner, deadline.
- Clarify and plan: define scope, acceptance criteria, and task slices.
- Implement: ship the smallest working change.
- Review gate: independent compliance and quality review.
- QA and debug loop: reproduce, diagnose, fix, re-verify.
- Learn: save reusable facts, patterns, and conventions.
Use z when you want:
- founder-led execution without a large agent org chart;
- clear handoffs between planning, implementation, and approval;
- an agent team that can scale from startup work to more regulated delivery;
- explicit checks against scope creep and low-signal AI output.
From a plugin-compatible client, point installation to this folder:
z/
Or copy the contents into a plugin source repository and install from source.
z/
plugin.json
apm.yml
README.md
agents/
z-lead.agent.md
z-planner.agent.md
z-implementer.agent.md
z-reviewer.agent.md
z-qa-debugger.agent.md
skills/
z/
SKILL.md
references/
INTAKE.md
PLAN.md
REVIEW.md
LEARNINGS.md
The skill ships with four lightweight reference templates:
- skills/z/references/INTAKE.md for task intake.
- skills/z/references/PLAN.md for compact planning.
- skills/z/references/REVIEW.md for independent approval.
- skills/z/references/LEARNINGS.md for post-task knowledge capture.