Where this fits

You've just finished a detailed implementation plan for a feature. The plan looks solid in your head, but the last time you skipped a review, the build ran headfirst into an unconsidered dependency and burned a sprint. Before implementation starts, you need a quick, repeatable way to check the plan against its own history and context—without leaving your agent's loop.

gza-plan-review wraps that check into an interactive skill: it inspects a plan task with built-in commands, surfaces the final output, and asks you to confirm the go/no-go signal before any code is written.

Why agents benefit

• Enforces a gate before costly work – The skill forces a structured review step, so an agent can't silently proceed to implementation with an unverified plan.
• Uses task metadata directly – It reads status, type, and history from your existing gza tooling, giving you real context instead of a generic checklist.
• Asks for human confirmation – Via AskUserQuestion, it makes the quality gate interactive, so the developer keeps final control over the go/no-go call.
• Surfaces failure context – If a plan failed previously, the skill shows the logs and history so you can decide whether the new plan addresses the root cause.
• Stops early on dead ends – When the task is missing or isn't a plan, it returns a no-go with corrective actions, preventing wasted cycles.

Practical scenarios

Pre-implementation sanity check on a complex refactor

You're about to refactor a legacy module. The plan task gza-1042 is marked complete, but you're unsure if it accounted for the recent schema migration. Run the skill, and it pulls gza show, gza log, and recent plan history. You review the final output, spot a missed foreign key, and issue a no-go with a note to update the plan before implementation.

Reusing precedent from past plan reviews

Your team has a pattern of planning new endpoints. The skill's history review shows that plans with a certain structure always required a second review. The agent spots that recurring pattern and flags it before you even open the plan, saving a round of feedback.

Gatekeeping a plan with incomplete dependencies

The plan references a component that hasn't been built yet. The skill's log check shows a failing integration test from last week. You see the context, decide to wait, and the skill returns a no-go—preventing a premature start.

Add it to your agent workflow

Install the skill directly from the CLI:

npx skills add mhawthorne/gza --skill gza-plan-review

Then invoke it by passing the prefixed plan task ID:

User: Review plan gza-1234 before we start
Agent: (runs gza show, log, and history) The plan is complete and clean. Do you want to give a Go?
User: Yes

For a detailed skill breakdown, see the OpenAgentSkill listing.

Compare before adopting

Evaluate the skill against your own quality signals: does it integrate with your existing plan/task format? Check maintenance freshness—the repo hasn’t been updated in 6 months, but the gza commands are stable. Compare it with other gatekeeping skills: does it allow interactive confirmation or is it fully automated? Ensure it matches your workflow’s need for human oversight.

Why it is worth tracking

Although the project has modest traction (11 stars), it addresses a real pain point: preventing premature implementation on flawed plans. Its focused scope and interactive design make it a candidate for teams already using the gza CLI. Track it if you want to codify quality gates without building your own review logic from scratch. When the repo shows new commits or community issues, it’s worth a deeper look.