do

Routes ambiguous and multi-part requests to DoFlow skills with task decomposition and effort estimation.

Updated Jul 9, 2026
One-click install
npx skills add https://github.com/khoavu882/do-flow --skill do-khoavu882
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: do
Source: https://github.com/khoavu882/do-flow/tree/main/core/shared/skills/do
Command: npx skills add https://github.com/khoavu882/do-flow --skill do-khoavu882

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? When a request does not clearly map to a single workflow, or bundles several unrelated asks across files and domains, it is unclear which command should handle it. This Skill acts as the DoFlow dispatcher: it announces session status, decomposes multi-part requests into validated work packages, routes each package to the right /do-* skill, selects the best available search or analysis tool, and produces scoped effort estimates before any work begins. ## Core Features & Use Cases - Command Dispatch & Session Status: With no arguments it announces the DoFlow session and shows the core command reference; with a known /do-* command it forwards directly to that skill. - Multi-Part Task Decomposition: Splits bundled requests into sequenced or independent work packages, proposes a task class per package, and has the runtime validate each class before routing (depths: shallow, normal, deep). - Capability & Tool Routing: Queries the live capability router to pick the right provider for search, dependency tracing, blast-radius analysis, or test execution, with fallback chains when Semble or Graphify are unavailable. - Effort Estimation: Produces range-based estimates anchored in git history and file scope, stating the assumptions that set the range width. - Use Case: A user asks "first refactor the auth module, then document the API, then review the changes." The Skill decomposes this into three packages, validates a task class for each, and routes them to do-implement, do-document, and do-code-review in sequence. ## Quick Start Ask the assistant to break down and route a multi-part request, for example: "Use /do to plan out fixing the login bug, updating the docs, and reviewing the changes."

Frequently Asked Questions about do

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I break down a multi-part coding request into separate tasks?▼

Invoke /do with the full request and it decomposes it into work packages, proposing one task class per package that the runtime validates. Use --depth shallow, normal, or deep to control decomposition granularity from a flat table to a full Epic-Story-Task hierarchy.

How do I choose the right tool for searching a codebase?▼

Run doflow capabilities or doflow route --intent <intent> to query the capability router, which reports which provider is actually available on your machine. Semantic search prefers Semble and falls back to Ripgrep, while dependency tracing prefers Graphify.

How do I get a development effort estimate before starting work?▼

Invoke /do with the --estimate flag and describe the change. It produces a range anchored against git history and file scope, stating the assumptions that set the range width, and stops without making any edits.

What happens when the runtime rejects a proposed task class?▼

The Skill stops and prints the rejection message verbatim, which lists valid classes and suggestions. It asks you to choose from the valid classes and re-validates, never substituting a default class on its own.

Does the do dispatcher modify code directly?▼

No. It only announces status, routes requests, validates task classes, selects tools, and produces estimates. Actual code edits are delegated to specialist skills such as do-implement, do-document, or do-code-review.