nazgul

Connect Nazgul task tracking to an external project board (GitHub Projects, Azure DevOps, etc). Use when user says "connect to github projects", "set up board", "track on github", or "nazgul board".

Start or resume a Nazgul autonomous development loop. Use when user says "start nazgul", "run nazgul", "begin development", "resume the loop", or passes an objective for new work. Auto-detects project state — no arguments needed.

Check the current state of a Nazgul autonomous loop. Use when asked about loop progress, task status, iteration count, review board status, or how the Nazgul loop is going.

Run one Nazgul automation-heartbeat tick — triages the work inbox and auto-starts the next objective if idle. Opt-in and default-off; fired by an optional Claude Code native scheduled agent (routine) or run by hand. Use when asked to "run a heartbeat tick", "check the inbox", or to test/debug the heartbeat.

View Nazgul run history — iteration timeline, task completions, review verdicts, git commits. Use after an overnight run to see what happened.

Brainstorm a new idea/objective into a Nazgul spec and tasks, then optionally run it. Interactive design front-end — produces a per-idea spec and a ready-to-execute task plan.

Generate a portable, Nazgul-free project bundle (docs + Claude subagents) without installing Nazgul. Runs the full pre-planning pipeline (discovery, doc-generator, reviewer-instantiation, optional designer) and emits output into standard paths (./docs/, ./docs/context/, ./.claude/agents/, ./.claude/).

Fully remove Nazgul from a project — deletes all runtime state, generated agents, MCP config, permissions, CLAUDE.md injections, and gitignore entries. Use when user says "remove nazgul", "uninstall nazgul", "clean nazgul", or wants to completely undo /nazgul:init.

View and change Nazgul settings — model assignments, formatter, notifications. Use when user says "configure nazgul", "change models", "nazgul settings", or wants to adjust config after init.

Collect targeted codebase context for a specific objective type. Use before planning to deeply understand the code that will be affected.

Scan the codebase to build a project profile and generate tailored reviewer agents. Use when setting up Nazgul for a new project, after major codebase changes, or when reviewer agents need updating.

View, generate, or regenerate project documents (PRD, TRD, ADRs, test plans). Use when asked about project documentation, design decisions, or technical specs.

Research latest Claude Code features and propose Nazgul improvements. Use periodically to keep Nazgul aligned with new platform capabilities. Run with `/loop 2w /nazgul:enhance` for auto-recurring checks.

Interactively build a project specification. Guides you through tiered questions — quick product overview first, optional deep-dive into user stories and success metrics. Outputs nazgul/context/project-spec.md.

Show Nazgul quick reference — all commands, modes, and getting started guide. Use when user says "nazgul help", "what commands", or needs orientation.

Initialize Nazgul for a project — check prerequisites, run discovery, create runtime directories, generate reviewer agents. Use when setting up Nazgul for the first time, user says "initialize nazgul", "set up nazgul", or before running any other Nazgul commands.

Distill recurring mistakes into numbered, human-approved Learned Rules. Runs the learner agent, then presents each candidate for approve/edit/reject. Supports --dry-run and --retire.

View loop performance metrics — task velocity, approval rates, retry distribution, reviewer stats. Use when asked about loop performance, development metrics, or how the loop is doing.

Lightweight task mode for bug fixes, config changes, and small features. Supports --no-review and --discuss flags.

Gracefully pause the Nazgul autonomous loop at the next iteration boundary. Use when user says "pause nazgul", "stop the loop", "halt nazgul", or wants to pause work without losing progress.

Reset Nazgul state to a clean slate. Archives current state and recreates from templates. Use when Nazgul gets into a corrupted or confusing state.

Manually trigger a review cycle for a specific task or the current IN_REVIEW task. Use when asked to review a task, run reviewers, or check review status.

Run a cleanup and simplification pass on all files modified during a Nazgul loop. Use after a Nazgul loop completes to improve code clarity without changing functionality.

Task lifecycle management — skip, unblock, add, prioritize, info, and list tasks. Use when you need to manage individual tasks in the Nazgul pipeline.

Human acceptance testing — structured verification that work actually works. Run standalone or integrated in HITL review cycle.