How michaelshimeles/skills works
This repository packages seven on-demand Claude Code skills rather than one application. The documented workflow starts with an isolated worktree, applies service-layer guidance during implementation, gathers evidence, then prepares visual proof and iterates Greptile review; prose cleanup applies throughout. Evidence-driven testing also includes a Python recorder. Other skill internals are represented only at the documented capability level because their source instructions were not sampled.
Main entry points: evidence-driven-testing/SKILL.md Test Evidence · before-and-after/SKILL.md Visual Comparison · new-feature/SKILL.md Isolate Work · code-structure/SKILL.md Structure Code
Components 12
Workflow Guidance
- Isolate Work
new-feature/SKILL.md - Structure Code
code-structure/SKILL.md
Evidence and Visual Proof
- Test Evidence
evidence-driven-testing/SKILL.md - Evidence Recorder
evidence-driven-testing/scripts/evidence.py - Visual Comparison
before-and-after/SKILL.md
Review and Writing
- Greptile Review
greploop/SKILL.md - Large-PR Review
greploop-apps/SKILL.md - Polish Prose
unslop/SKILL.md
Other components
- Developer
- Changed App
- PR and Tracker
- Greptile
Connections 12
- Developer → Isolate Work: starts with
README.md - Isolate Work → Structure Code: then builds
README.md - Structure Code → Test Evidence: then proves
README.md - Test Evidence → Evidence Recorder: uses recorder
README.md - Test Evidence → Changed App: tests live
README.md - Test Evidence → PR and Tracker: posts evidence
README.md - Test Evidence → Visual Comparison: then ships
README.md - Visual Comparison → PR and Tracker: shares comparison
README.md - Visual Comparison → Greptile Review: then reviews
README.md - Greptile Review → Greptile: triggers review
README.md - Large-PR Review → Greptile: tags app
README.md - Developer → Polish Prose: polishes writing
README.md