From maverick
Implements a focused code change as part of the Maverick workflow, applying coding standards and running preflight checks. Designed for one task per invocation from within issue or epic phases.
How this skill is triggered — by the user, by Claude, or both
Slash command
/maverick:do-code brief description of the task (one line)brief description of the task (one line)The summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Depends on:** mav-bp-error-handling, mav-bp-logging, mav-bp-application-security, mav-local-verification, mav-scope-boundaries, mav-systematic-debugging
Depends on: mav-bp-error-handling, mav-bp-logging, mav-bp-application-security, mav-local-verification, mav-scope-boundaries, mav-systematic-debugging
Wraps implementation work in a Maverick skill so the workflow report
records what was done and the agent applies the project's coding
standards before editing. Invoke this skill once per task from
inside do-issue-solo / do-issue-guided /
do-epic Phase 5. The argument is a one-line task
description that gets logged in the report.
This skill is a thin wrapper. The actual edits are made by the orchestrating Claude Code session using its built-in tools (Read, Edit, Write, Bash, Grep). The skill's job is to pin standards and scope, not to perform the edits in a subagent.
Before doing anything else, check for repo-specific coding guidance at
docs/maverick/skills/do-code/SKILL.md (path relative to the repo root).
Run this first. If it exits non-zero, halt and report the stderr output to the user verbatim. Do not proceed.
uv run maverick preflight do-code
The check verifies the project is initialised and uv is on PATH.
Before editing, read and follow:
mav-scope-boundaries — implement only what ``
asks for. No unrelated cleanup, no opportunistic refactors, no
abstractions for hypothetical future requirements. Three similar
lines is better than a premature abstraction.mav-bp-error-handling — error handling at system
boundaries only (user input, external APIs). Don't catch what
framework guarantees rule out. No defensive code for scenarios that
can't happen.mav-bp-logging — log levels, structured fields, and
the project's logging conventions.mav-bp-application-security — input validation, secret
handling, authn/authz patterns. If the change touches any of these,
surface the design choice in your verification.Also read any project-level skills at docs/maverick/skills/ —
topic-specific guidance for this codebase that supplements the
maverick-wide best practices.
mav-local-verification (lint,
typecheck, tests). Run the project's verification commands.mav-systematic-debugging and fix. Do not paper over
a failing test or silence a lint warning. Do not commit red.git push --force,
git reset --hard, rm -rf, drop tables, or delete data. If the
task description doesn't explicitly authorise a destructive step,
treat it as out of scope and ask.do-test as a sibling call before the commit).do-code is a subroutine of the calling workflow's
per-task loop — not a terminal action. Returning from this skill is a
hand-back to the caller's next numbered step, not a completion event
(#106).
When you return from this skill, do not post a closing summary, do not stop, do not treat verification-green as "task done." The calling workflow's loop still owns, in order:
skill-dispatch interval that wrapped this invocation
(uv run maverick report end skill-dispatch … --outcome success).do-test for tests that
weren't folded into this skill.uv run maverick report commit …).If you find yourself drafting a final summary after returning here,
that is the signal: scroll back to the calling workflow's per-task
loop and resume from the step immediately after the
/do-code invocation.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub thermiteau/maverick --plugin maverick