Audit code quality by running parallel checks against a diff or project tree.
npx skills add https://github.com/LarryHsiao/Skadi --skill nazgul
Please help me install this Agent Skill. Skill: nazgul Source: https://github.com/LarryHsiao/Skadi/tree/main/skills/nazgul Command: npx skills add https://github.com/LarryHsiao/Skadi --skill nazgul
Nazgul provides an auditable, non-destructive way to enforce code quality and policy conformance by dispatching multiple checks in parallel against the chosen target (diff or project) and summarizing results in a single table.
Invoke /nazgul to start the audit on the current target.
High-intent search queries and answers about installing and using this skill.
To audit code quality in parallel, define checks with frontmatter fields in a checks/ directory and execute them against the diff scope. Nazgul dispatches all checks simultaneously and aggregates pass, fail, or n/a verdicts into a single summary table.
Yes, you can enforce governance policies non-destructively by running project audits. The system evaluates checks against the project tree or diff and returns verdicts without performing changes, commits, or filing issues.
The best way to run multiple code quality checks on uncommitted changes is using the diff scope. This evaluates all defined checks concurrently against the current diff and returns a concise table of pass, fail, or n/a results.
Parallel code audits work by dispatching custom check definitions simultaneously. Each check requires specific frontmatter fields including name, scope, agent, and autofix, ensuring deterministic execution and aggregated verdict reporting.
Yes, the project audit tool supports evaluating specific commit ranges through the diff scope. You can target uncommitted changes, specific commits, or commit ranges, executing all checks in parallel for consistent governance.
A key limitation of non-destructive code quality audits is that they never perform changes, commits, or file issues. The system strictly evaluates and aggregates pass, fail, or n/a verdicts into a table without modifying the project tree.
UsernameTronDefine a complete YAML frontmatter schema for Claude Code subagent .md files.
amiable-devEvaluate pull requests with LLM Council and produce pass/fail/unclear verdicts.
LumenWipeReview code diffs adversarially and triage findings into actionable categories.
plurigridLint Clojure code with clj-kondo and classify diagnostics using Gay.jl 3-color.
