vibe-coding-workflow

name	vibe-coding-workflow
description	Use when applying the .ai CORE_SKILL.md operating system, choosing Copilot CLI or VS Code playbooks, doing inspect-first coding, minimal patches, debugging, repo edits, docs, beginner handoffs, verification, or reusable AI coding workflows.
argument-hint	Describe the task and, optionally, the playbook to apply: debugging, edit-existing-repo, build-from-scratch, docs-and-specs, vscode-agents, copilot-cli, beginner-handoff.

Vibe Coding Workflow

Use this skill to apply the repository's .ai operating pattern: core behavior first, one job-specific playbook second, and the user's task third.

The goal is useful, verifiable progress with small changes, explicit assumptions, and a clear handoff.

Source Files

• Core operating rules: .ai/CORE_SKILL.md
• Usage guide: .ai/COPILOT_CLI_AND_VSCODE_GUIDE.md
• Playbooks: .ai/playbooks/

When to Use

• The user asks to use CORE_SKILL.md, a vibe coding workflow, Copilot CLI guidance, or VS Code agent guidance.
• The task benefits from an inspect-first, patch-minimally, verify-afterwards workflow.
• The user wants a reusable prompt, coding routine, playbook stack, beginner handoff, debugging process, docs process, or repo-editing process.
• The task is vague enough that the real objective, assumptions, constraints, and verification method need to be made explicit.

Workflow

1. Restate the real objective in one or two lines.
2. Identify material assumptions, constraints, dependencies, and likely failure points.
3. Choose the smallest useful playbook stack:
   • Always apply .ai/CORE_SKILL.md.
   • Add one main playbook from .ai/playbooks/ based on the job.
   • Add a second playbook only when it clearly complements the first.
4. Inspect relevant files or outputs before proposing code changes.
5. Make a short plan with verification points.
6. Implement the smallest viable change that maps directly to the task.
7. Verify with the strongest repo-native check available.
8. Hand back the result with what changed, why, what was verified, what remains unknown, and the next best step.

Playbook Selection

• debugging.md: something is broken; reproduce or inspect the failure mechanism before patching.
• edit-existing-repo.md: patching a live codebase; preserve architecture and touch only required files.
• build-from-scratch.md: starting a new project; make the smallest end-to-end version first.
• docs-and-specs.md: writing guides, specs, plans, prompts, or instructions.
• beginner-handoff.md: output should be plain-language, low-friction, and usable by a non-expert.
• vscode-agents.md: working in VS Code agent mode with workspace tools and file edits.
• copilot-cli.md: terminal-first Copilot CLI sessions or reusable CLI prompt patterns.

Decision Rules

• Prefer mechanism over slogans: explain what sustains the result, what it depends on, and what can fail.
• Prefer small, reversible edits over broad rewrites.
• Surface ambiguity when it materially changes the outcome; otherwise choose a conservative default and proceed.
• Mention simpler approaches when they exist.
• Do not add speculative features, unused abstractions, or future-proofing that makes the current task harder.
• Match the existing codebase style and leave unrelated issues as follow-up notes.

Quality Checks

Before finishing, confirm:

• The work maps directly to the user's request.
• Assumptions and trade-offs are explicit where they matter.
• The chosen playbook stack is appropriate and not overloaded.
• Verification was run or the reason it could not be run is clear.
• The final handoff is usable by the target reader, especially if the user asked for beginner-friendly output.

Reusable Prompt Pattern

Use or adapt this when the user wants a portable prompt:

Use `.ai/CORE_SKILL.md` as the core operating instructions and apply `.ai/playbooks/[chosen-playbook].md` for this task. First inspect the relevant files, restate the real objective, list material assumptions, make the smallest viable change, verify with the strongest repo-native checks available, and finish with a plain-English handoff that says what changed, what was verified, and what remains unknown.