frictionlesscode

Claude Workflows — Productivity skill for Claude Code

Productivity community

Claude Code plugin: two workflow skills layered on Superpowers.

How to install Claude Workflows

This entry records only its repository, not the path inside it, so there is no exact command to give. Open frictionlesscode/claude-workflows and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Claude Workflows does

Claude Code plugin: two workflow skills layered on Superpowers. fixing-raw-bugs triages unstructured bug reports into evidence-backed, cost-tiered batches; writing-technical-articles drafts long-form pieces without inventing facts.

Alternatives in Productivity

  • Cs2 Technical Analysis — You are an Expert 3D Technical Director specialized in CS2 (Counter-Strike 2) weapon and item pipelines 14.1k ★
  • Tradingview MCP — Real-time crypto & stock screening, advanced technical indicators, Bollinger Bands intelligence, candlestick p 2.6k ★
  • Repurpose — Workflow recipe — turn one blog post, video, or idea into a full platform-native content pack (thread, LinkedI 1.3k ★

README

claude-workflows

Two Claude Code skills I use on my own projects. Both layer on top of [Superpowers](https://github.com/obra/superpowers) and both refuse to run without it rather than falling back to a worse version of what it already does.

Install

/plugin marketplace add frictionlesscode/claude-workflows
/plugin install claude-workflows@frictionlesscode

Superpowers first, if you do not already have it:

/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace

What is in here

`fixing-raw-bugs`

An intake layer for the pile that arrives before a plan exists. Superpowers is excellent from `here is what I want built` onward and has no opinion on twenty unsorted field reports, some of which are feature requests and some of which cannot be reproduced at all with the tooling you have today.

The work is in three reference files:

  • references/triage.md turns a pile into an ordered list. Cheap read-only agents fan out one per bug and report facts only, never conclusions, because a model that cheap asked how hard a bug is will guess low and sound certain. Tiering happens centrally afterward, nothing is mechanical without a file:line, and unknown escalates.
  • references/decisions.md separates a code problem from a product problem. Design questions get asked in one batch, all of them rather than the two or three that look important, each with concrete options and a real example. Chat is the default; file mode is opt-in and sticky.
  • references/execution.md is the part that saves money. Batch composition, waves that share a surface, local concurrency limits, and the tier-to-model mapping.

Use `/fix-raw-bugs`, or just describe a bug backlog and it triggers on its own. It writes `bugs/STATE.md` and resumes from whatever phase that names, so a second run continues rather than re-triaging.

**You will probably need to build an instrument first.** Superpowers w