Plugin commands

This page lists everything in version 0.16.0 of the anthara plugin. For how the plugin loads and what its hooks do, see The anthara plugin.

Type a command in Claude Code with the /anthara: prefix, for example /anthara:start.

Commands you run

Decide and plan

Command	Purpose	You can pass
/anthara:start	Classifies the work and suggests which command to use. It asks before starting anything.	A description of the work, or a link to a brief or spec
/anthara:brainstorm	Generates and compares options for an open-ended problem.	The problem or topic
/anthara:discovery	Combines research from many sources into one brief that names each source.	A file, a document link, or pasted text
/anthara:spec-writer	Writes a spec split into slices. It adds the requirements from the packs that apply.	A brief, a spec, or a file path
/anthara:create-ticket	Creates a ticket in Jira, Linear, GitHub Issues, or Azure DevOps after you confirm a preview.	A spec, or nothing to describe the ticket yourself
/anthara:planner	Writes the build plan for a spec, a ticket, or a prompt, and saves it under docs/plans/.	A spec path, a ticket, or a prompt

Build and fix

Command	Purpose	You can pass
/anthara:plan-implementer	Builds a plan one slice at a time, with code and tests together. It reviews the whole change at the end.	A plan path, or nothing to use the newest plan
/anthara:debug-fix	Reproduces a bug, tests one cause at a time, and lands the fix behind a failing test.	The symptom, an error message, or a failing test
/anthara:collaboration-loop	Finds the > @anthara comments in a spec or brief, makes the changes, and replies inline.	A document link or file path
/anthara:workflow-builder	Interviews you and writes an agentic workflow YAML file. It writes connection names and never credentials.	A path for the YAML file

Review

Command	Purpose	You can pass
/anthara:code-review	Reviews a pull request from several angles and scores each finding from 0 to 100. It posts findings that score 80 or more as one pull request comment.	A pull request number or link, and --min-confidence N
/anthara:pr-walkthrough	Explains a pull request as a short guide, in the terminal and as a local HTML page. It explains and asks questions. It does not judge.	A pull request number or link

Keep up

Command	Purpose	You can pass
/anthara:inbox	Collects the work waiting for you in your connected tools and ranks it in one table.	A hint, such as “on-call today”
/anthara:team-update	Summarizes what your teammates did. The default period is the last 18 hours.	A period, such as 3d or since monday

/anthara:inbox and /anthara:team-update read from whichever MCP servers you have connected in Claude Code. Anthara does not supply those connections.

Skills that load automatically

These eight skills hold guidance. Claude loads them when the work calls for them, and other commands and agents load them too. You don’t need to run them.

Skill	What it covers
code-craft	Writing code: naming, function design, error handling.
test-craft	Writing tests: independence, meaningful assertions, reusing existing tests.
review-craft	Reviewing changes: what to check and how severe a finding is.
architecture-craft	Choosing an architecture style and keeping dependencies pointed the right way.
design-craft	UI and UX decisions, including accessibility minimums.
ai-ergonomics	How well code supports AI-assisted work, such as file size and explicit types.
test-execution	When to run which tests during a build.
browser-verification	Checking a running app in a real browser.

Claude Code also lists these eight under /anthara:, because the plugin does not hide them. Running one only loads its guidance.

Agents

Agents are helpers that a command starts for one job. You don’t start them yourself.

Agent	Started by	Job
code-orienteer	planner, plan-implementer	Reads the code a spec touches and writes down the patterns to follow.
slice-builder	plan-implementer	Builds one slice: code, tests, and a commit.
reviewer	plan-implementer	Reviews the finished change for quality.
pack-reviewer	plan-implementer	Reviews the finished change against one pack’s rules. One runs per pack.
browser-verifier	plan-implementer	Checks UI slices in a running app.
spec-heal	plan-implementer	Writes an agreed change of approach back into the spec and plan.
recap	plan-implementer	Walks you through what was built.
hypothesis-prober	debug-fix	Tests one possible cause of a bug, read-only.
root-cause-auditor	debug-fix	Challenges the stated cause before the fix is written.