Agentic Coding Flywheel Skills Bundle

A collection of Claude Code skills extracted from Jeffrey Emanuel's Agentic Coding Flywheel methodology. These skills capture the key workflows from the 9-step process for building software with AI agent swarms.

Skills

/deep-planning - 85% Planning First

Enforce comprehensive upfront planning before any implementation. Produces a markdown plan covering architecture, data flow, error handling, edge cases, and security. Based on the principle that plan mistakes cost 1x to fix, while code mistakes cost 25x.

/competing-plans - Parallel Plan Synthesis

Spawn 3 parallel agents that independently plan the same feature, then synthesize a "best-of-all-worlds" plan. Each agent has a different persona (Pragmatist, Defender, Architect) that emphasizes different concerns. The synthesis combines the strongest elements from each.

/idea-wizard - Structured Feature Ideation

A divergent-convergent ideation pipeline: generate 30 ideas grounded in the existing codebase, winnow to 5 best, re-expand to 15 refined ideas around the strongest themes. Avoids both tunnel vision and unfocused brainstorming.

/bead-polishing - Iterative Task Refinement

Run 4-6 polishing rounds on a task breakdown until convergence. Each round: deduplicate, verify coverage against the plan, fix dependencies, fill gaps, check for oversimplification. Stop at ~75% convergence. Works with any task format (beads, GitHub issues, PRD stories).

/multi-mode-review - Three-Mode Code Review

Three review modes that catch different bug categories:

• Fresh Eyes: Self-review for logic errors, edge cases, off-by-ones
• Cross-Agent: Independent agent reviews for integration/boundary issues
• Random Exploration: Trace execution flows for architectural drift and systemic bugs

Recommended Flow

/idea-wizard               "What should I build?"
      |
/deep-planning             Comprehensive markdown plan (85% of effort)
      |
/competing-plans           3 parallel plans -> best-of-all-worlds synthesis
      |
/bead-polishing-ralph      Create beads (if none exist) + 4-6 polish rounds
      |                    until ~75% convergence. Enforces ralph-tui constraints:
      |                    one-context-window sizing, quality gates, completion signal.
      |
[implementation]           ralph-tui run --tracker beads-rust
      |
/multi-mode-review         Fresh Eyes + Cross-Agent + Random Exploration

Note: /bead-polishing-ralph can create beads from the plan if none exist yet, or you can use /ralph-tui-create-beads first and then polish separately. /bead-polishing is the generic (non-ralph) variant for GitHub issues or other formats.

Installation

Copy the skill directories into your Claude Code skills directory:

cp -r skills/* ~/.claude/skills/

Or symlink them:

for skill in skills/*/; do
  ln -s "$(pwd)/$skill" ~/.claude/skills/$(basename "$skill")
done

Source

Based on The Complete Flywheel Guide by Jeffrey Emanuel.

The original methodology covers a 9-step workflow for building software with coordinated agent swarms, including concepts like beads (self-contained work units), Agent Mail (inter-agent communication), and convergence-based refinement. These skills extract the most actionable workflows into Claude Code slash commands.

License

MIT