do-it

English | 中文

Professional coding judgment for Codex, Claude Code, Cursor, Pi, and Grok Build. do-it supplies task-fit skills, independent specialists, compact context, and useful deterministic checks. The model chooses methods.

Core stays small: preserve the user's goal, settled decisions, and authorization boundary; work from current facts and distinguish assumptions; fix the causal owner with necessary scope; support claims with relevant evidence and name gaps. Its self-check questions are optional thinking aids, not a required interview, workflow, or report.

Install

Host	Command
Codex	codex plugin marketplace add tdwhere123/do-it && codex plugin add do-it@tdwhere-do-it
Claude Code	/plugin marketplace add tdwhere123/do-it → /plugin install do-it@do-it
Cursor	npm run install:cursor-local → Reload Window
Pi	pi install npm:@tdwhere/do-it-pi
Grok Build	Generated bundle installation

Installation details cover host setup and smoke checks. The 0.18.1 source update is unreleased; marketplace or npm installation may serve an earlier version. Codex specialists require the separate native-agent setup described in that guide. OpenCode and Kimi Code repository support is retired; local applications and user configuration remain untouched. Talk normally after installation. No task tier, mandatory router entry, or .do-it scaffold is required. Reuse existing instructions and project documents; write lightweight task notes only when they help continuity.

Professional perspectives

Choose a skill directly when its description helps the work:

Skill	Useful judgment
do-it-core	Stable principles for intent, facts, causal scope, and claims
do-it-code-quality	Causal ownership, contract fallout, and stateful failures
do-it-architecture	Authority, boundaries, compatibility, migration, and recovery
do-it-decide	Decision-changing uncertainty and meaningful alternatives
do-it-review	Requirements and implementation quality as distinct lenses
do-it-audit	Explicit deep audits with file, layer, and chain coverage and independent causal synthesis
do-it-verify	Relevant evidence and the practical limits of delivery claims
do-it-context	Project terminology, invariants, and stable knowledge worth preserving in existing docs
do-it-retrospective	On-demand reflection on observed outcomes
do-it-skill-authoring	Succinct triggers, unique judgment, and necessary boundaries

do-it-router remains only as an explicit compatibility alias for discovery.

Independent agent contexts matter: they can gather fresh evidence and reach conclusions without inheriting the parent's anchoring. Pass goals, scope, settled decisions, and source facts; label parent opinions as hypotheses. Meaningful specialist roles retain read-only or scoped-write restrictions. The parent integrates results and verifies the combined work. Use independence when it adds concrete value, with no blanket worker quota or forced delegation stage.

Deep audit on request