Skip to content
Vaizer
โ† Skills
๐ŸŸฃ Nekko officialCoding

nyaa

by Nekko Labs

Five specialist reviewers read your pull request or working diff, each hunting one kind of problem, and their findings merge into a single verdict. The heaviest one reads intent before code: your PR description says what this change is for and what it is worth, your SPEC.md says what the product is, and the review is whether the diff actually delivered that.

  • Holds the code to what the PR said it would do. It reads the description first, pulls out the goal and the value being claimed, then traces that claim through the diff. Not met, or claimed and missing, is blocking. An empty description is a finding too, and it drafts one for you.
  • Checks the change against your spec. The heaviest lens reads SPEC.md before the code, then flags anything the spec never promised and any feature that shipped without the spec being updated. When the change contradicts a feature the spec still describes, it says so as a stale spec, with the sentence to change.
  • Reviews the spec itself, not just the code. Missing feature definitions, no stated test expectation, undocumented conventions. Each gap comes back with the sentence to add, so the spec improves instead of just being blamed.
  • Four more lenses, one concern each. Security, dependencies and supply chain, correctness and concurrency, style and lint. One reviewer per concern stays sharper than one reviewer for everything.
  • Runs your real tools, not just the diff text. The dependency cat runs your audit; the style cat runs your linter and separates failures this change caused from ones that were already there.
  • You choose who reviews. All five sit by default. Narrow the council with --cats, --skip, or --pick for an interactive roster. The verdict always names who sat, so a partial pass can never read as a clean one.
  • Folds in the bots you already run. Codex and Dependabot findings are pulled in and cited, including when Codex skipped the PR, which is precisely when you need to know nothing else reviewed it.
  • code-review
  • pull-request
  • pr-description
  • spec
  • spec-conformance
  • product-intent
  • security
  • dependencies
  • supply-chain
  • concurrency
  • lint

How it works

The steps this skill runs through when you invoke it. Click any step to see what it does.

Workflow

TriggerContextAgentToolDecisionLoopOutput

Get this skill

1. Add the marketplace (once)

/plugin marketplace add nekko-labs/vaizer

2. Install the skill

/plugin install vaizer@vaizer
View source

Skills run with your machine's permissions. Read the source before you run anything you didn't write.


Was this skill helpful?

Understand your own agents next

Break down any public skill into a workflow you can read, or watch a long-running loop make its way toward the goal.