mcpField
E
explain-workflow

explain-workflow

github.com/klausborgesAgent workflows1112 installsUpdated 1w ago
Repository

About explain-workflow

explain-workflow is a skill that helps users choose and understand a workflow for software tasks. It identifies a relevant skill, gives an exact next prompt, and describes what to expect without starting the work.

Key features

  1. 01Maps requests to relevant workflow skills
  2. 02Provides an exact next prompt
  3. 03Explains expected outcomes
  4. 04Distinguishes planning, implementation, diagnosis, and review

Use cases

  1. 01Choosing a workflow for a feature request
  2. 02Finding the next step for a failure investigation
  3. 03Selecting a review or handoff workflow

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/klausborges/workflow-skills --skill explain-workflow

After installing, type/explain-workflowin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

skill.toml
agents/
openai.yaml
references/
workflow-language.md

Details

From the project README

Explain Workflow

Explain the workflow from the user's point of view. Use workflow-language.md.

This skill is a guide, not a router. Give the relevant skill, an exact next prompt, and what to expect; do not start the work unless asked.

Invocation

  • Explicit: use plan-work to plan this feature or use review-work to review this diff.
  • Natural: plan this feature, diagnose this failure, write a handoff, or review this change.
  • Correction: name the skill when the agent chose the wrong workflow.

Recipes

  • Uncertain or high-risk work, or an explicit planning request -> plan-work. Planning can stay in chat when requested; otherwise a saved ephemeral Plan is the default.
  • Already-decided work from a Plan, handoff, findings, direct brief, docs task, fix, or refactor -> implement-work. Plan tracking applies only when a Plan governs the work; verification and independent fresh-context review always apply.
  • Tiny clear implementation -> work directly; it still gets a fresh-context review before delivery.
  • Code orientation -> explain-codebase.
  • Root-cause investigation -> diagnose-issue; fixes require explicit authorization.
  • Plan, implementation, continuation, or artifact review -> review-work; review is read-only unless fixes are separately requested.
  • Continuation prompt or file -> write-handoff; file/chat storage is separate from Immediate/Durable lifecycle.
  • Architecture opportunity discovery -> improve-architecture; concrete change review belongs to review-work.
  • Bounded behavior-preserving simplification -> simplify-work.
  • Disposable design experiment -> plan-prototype.
  • Unsure but substantial -> use-workflow.

Keep the answer short unless the user asks for the full map. If the question is “how do I use this for X?”, explain and give the next prompt rather than silently starting X.

Comments

Markdown supported