AI agent skills.

prep-dev

Examine the premise and clarify the implementation through repository research and a short interview.

fusion

Run two user-selected models on the same prompt, compare their evidence, and reconcile the answers.

unit-tests

Evaluate test relevance and protect business use cases and domain invariants in any programming language.

spec-gap

Compare an implementation with its product requirements. Identify missing behavior, coverage, and work beyond the spec.

sheriff

Review code changes and triage issues using evidence, severity, diagnostic confidence, and operational exposure.

simplify-code

Simplify working code so it is easier to read and maintain, while preserving its behavior.

prep-pr

Check scope, contracts, maintainability, and validation before preparing a pull request for review.

warden

Check intent, code review, tests, documentation, and lint before delivering changes within the authorized scope.

docs-writer

Write and improve developer documentation with runnable examples, progressive explanations, and precise references.

gatekeeper

Track acceptance criteria and verify the evidence for each deliverable before reporting a substantial task complete.

Adapted from Leonxlnx/unlazy.

memory-sql

Save and selectively recall project facts, decisions, tasks, and evidence across sessions using SQLite.