Agent workflow skills
Workflow tooling for product and engineering teams: reusable AI playbooks, review loops, and implementation briefs that turn vague ideas into shipping decisions.
Make the work easier to review.
A pull request needs more than a summary. Reviewers need to understand the intent, inspect the actual changes, and distinguish verified findings from plausible concerns. This toolkit covers both starting a change and explaining it once code exists.
04 / Technologies & context
- Python
- Developer Tools
- AI Agents
- Open Source
Inside the project
Implementation
Python standard-library scripts handle branch planning, diff parsing, and HTML rendering. Host-specific manifests package the same plugins for Codex, Claude Code, and Cursor. The walkthrough accounts for diff hunks in logical modules; the brief explains user-facing implications. Package validation checks the scripts and manifest wiring.
The workflow
- Start work in an isolated branch and worktree
- Review diff hunks with cited evidence
- Generate a readable walkthrough and product brief
Scope & status
GitHub writes are separate, explicit actions. A generated review is evidence to inspect, not a guarantee that the code is correct. This is a personal toolkit, distinct from Quivly Skills.