-
Notifications
You must be signed in to change notification settings - Fork 2
claude code
Pre-Alpha. This page describes behavior that may change.
Ze is developed with Claude Code, and the repo ships a set of project-specific slash commands under .claude/skills/. They are available inside Claude Code when you are working in the Ze repository and they encode the project's workflow: spec-driven, TDD, multi-agent review, structured commits. This page is the cheat sheet.
The canonical version is main/docs/claude-code-cheatsheet.md.
| Command | What it does |
|---|---|
/ze-status |
What needs my attention? Dashboard with suggested next actions. |
/ze-explore <topic> |
Find and read all files related to a topic. |
/ze-spec |
Create or resume a spec (a design document for a task). |
/ze-design |
Stress-test a design through structured decisions. |
/ze-audit |
Pre-implementation: what code already exists for this spec? |
/ze-implement |
Implement a spec end to end (TDD, review loops, docs). |
/ze-verify |
Run make ze-verify and format the results. |
/ze-debug |
Investigate a failing test with parallel hypotheses. |
/ze-review |
Quick single-pass review of uncommitted changes. |
/ze-review-deep |
Exhaustive review with nine specialised agents. |
/ze-review-spec |
Did we build what the spec says? |
/ze-review-docs |
Documentation accuracy and completeness. |
/ze-commit |
Prepare a scoped commit script (does not commit directly). |
/ze-recap |
Summarise the current session state. |
/ze-handoff |
Generate a handoff document for the next session. |
/ze-rfc |
Generate an implementation summary from an RFC. |
/ze-find-alloc |
Scan encoding paths for allocations that should use buffers. |
/ze-fix-alloc <file:line> |
Convert a specific allocation to buffer-writing. |
/ze-extract <src> <dst> <symbols> |
Move Go symbols between files. |
The standard loop, start to finish.
/ze-explore <area> # Understand what exists.
/ze-spec # Write the spec (design, acceptance criteria, test plan).
/ze-audit # Check what is already implemented.
/ze-implement # TDD implementation with built-in reviews.
/ze-verify # Run the full test suite.
/ze-review-deep # Multi-agent review before submission.
/ze-commit # Prepare the commit.
The maintainer runs the same workflow on every contribution before merging, so running it yourself before submitting is the fastest way to get a patch through review.
/ze-debug # Paste the failing test output, get parallel investigation.
/ze-verify # Confirm the fix does not break anything else.
/ze-review # Quick review of the fix.
/ze-commit # Prepare the commit.
Four different review commands for four different scopes.
/ze-review # Quick single-pass, uncommitted changes, ~2 minutes.
/ze-review-deep # Exhaustive, nine parallel agents.
/ze-review-deep branch # Same, but reviewing the whole branch vs main.
/ze-review-spec # Check implementation against spec.
/ze-review-docs # Check documentation against code.
/ze-status # What needs attention?
/ze-recap # Where am I in the current task?
/ze-handoff # Prepare state for a new session.
The four review commands are not redundant. They cover different scopes at different depths.
| Command | Scope | Depth | When to use |
|---|---|---|---|
/ze-review |
Uncommitted changes. | Single pass, ~2 min. | Quick sanity check. |
/ze-review-deep |
Uncommitted or branch. | 9 parallel specialised agents. | Before merge or before submitting a large change. |
/ze-review-spec |
Selected spec. | Spec vs code comparison. | After implementing a spec. |
/ze-review-docs |
docs/ directory. |
Accuracy and completeness. | Periodic documentation audit. |
/ze-review-deep is the one the maintainer runs on every contribution. Running it yourself first catches the problems the maintainer would have caught.
A few things to know about the slash commands.
/ze-commit does not run git commit. It generates a commit script that you run yourself. This is deliberate: commits are a human decision, and the tool makes it easy to prepare one without taking away your ability to inspect it first.
/ze-verify is the same as make ze-verify with structured output. It does not do anything the Make target does not do.
All review commands are read-only. They report findings, they do not make changes.
The commands assume a working tree. They do not push to a remote, they do not force-push, they do not rewrite history. Commits that need that treatment are for the human.
- Contributing for the surrounding contribution workflow.
- Coding standards for the rules the commands enforce.
-
main/docs/claude-code-cheatsheet.mdfor the canonical version.
Adapted from main/docs/claude-code-cheatsheet.md.
Unreviewed draft. This wiki was authored in bulk and has not been reviewed. File corrections on the issue tracker.
- Overview
- YANG Model
- Editor Workflow
- Archive and Rollback
- System
- Interfaces
- BFD
- FIB
- Firewall
- Traffic Control
- L2TP/PPP
- VPP Data Plane
- RPKI
- TACACS+ AAA
- Fleet
- BGP
- Starting and Stopping
- Show Commands
- Monitoring
- Logging
- Operational Reports
- Healthcheck
- MRT Analysis
- Upgrade and Restart
- Storage
- Policy
- Core
- Resilience
- Validation
- Capabilities
- Address Families
- Protocol
- Subsystems
- Infrastructure
- Route Server at an IXP
- Transit Edge with RPKI
- Public Looking Glass
- ExaBGP Migration Walkthrough
- FlowSpec Injection
- Chaos-Tested Peering
- AS Path Topology