prompt-pack

name	prompt-pack
description	Generates a self-contained prompt pack from a spec. Embeds real code patterns from .vibeflow/ so any coding agent (Claude Code, Cursor, Copilot) follows the project's conventions without needing repo context. Use when handing off implementation to a separate session or agent.
argument-hint	<spec file or feature>
allowed-tools	Read, Grep, Glob

---

Description and examples

What it does: Reads a spec from .vibeflow/specs/ (or path), builds a single prompt that includes objective, DoD, anti-scope, and real code patterns. Saves to .vibeflow/prompt-packs/<slug>.md. Give that file to the coding agent so it implements without needing the rest of the repo context.

Examples:

• /vibeflow:prompt-pack .vibeflow/specs/login-flow.md — Generate prompt pack from the login-flow spec.
• /vibeflow:prompt-pack login-flow — Same; agent finds the spec by name.

---

Language

The prompt pack MUST be written in the user's detected language. The opening line should be language-adaptive: Write the opening line in the user's detected language. The concept is: "You are only seeing this prompt; there is no context outside it."

All textual content (objective, DoD, anti-scope, guidance) should be in the detected language. Code, paths and technical names remain in English.

Generate a self-contained prompt pack for: $ARGUMENTS

Steps:

0. Validate spec size. Read the spec and check:

   • If it has >7 DoD checks, OR
   • If it exceeds the budget (from .vibeflow/index.md or default ≤ 6 files)

   If EITHER condition is true, stop. Do NOT generate the prompt pack. Instead, inform the user: "This spec exceeds limits (N DoD checks / N files). Run /vibeflow:gen-spec again to split it into smaller specs before generating a prompt pack."

1. If $ARGUMENTS is a file path, read that spec file.

2. If it's a feature description, look for a matching spec in .vibeflow/specs/.

3. If no spec exists, generate one first (following gen-spec format), save it, then continue.

4. Read .vibeflow/ docs:

   • .vibeflow/conventions.md (always)
   • Pattern docs listed in the spec's "Applicable Patterns" section (manual override)
   • If the spec doesn't list patterns: use Pattern Resolution. Read the ## Pattern Registry YAML block from index.md (between <!-- vibeflow:patterns:start/end --> markers). Cross-reference the registry's tags and modules against the spec's scope. Load the top 3-5 matching patterns. If no registry exists, infer which ones are relevant.

5. Read the actual codebase files relevant to this task.

6. Generate the prompt pack.

The Prompt Pack MUST Start With:

You are only seeing this prompt; there is no context outside it.

(Write this opening line in the user's detected language.)

Then include, in this order:

1. Objective and Definition of Done

Copied from the spec. Non-negotiable.

2. Anti-scope

What NOT to do. Copied from spec.

3. Budget

Max files to change. Read the budget from the spec. If the spec doesn't specify a budget, check .vibeflow/index.md for the Suggested budget line. If neither is available, default to ≤ 6.

4. Project Patterns to Follow

THIS IS CRITICAL. Embed the actual patterns the agent must follow. Copy the relevant sections from .vibeflow/patterns/*.md and .vibeflow/conventions.md directly into the prompt pack.

Include REAL code examples from the pattern docs — not references to external files the agent won't see. The prompt pack is self-contained.

Format as:

## Patterns to Follow

### <Pattern Name>
<description>
<real code example from this project showing the pattern>

### Coding Conventions
<relevant conventions for this task>

5. Where to Work

Real file paths from the codebase (verify they exist). Relevant code snippets for context.

6. Directional Guidance

Architectural direction, constraints to respect. NOT step-by-step prescriptive instructions.

7. How to Run/Test (MANDATORY)

Commands to validate the implementation. This section is REQUIRED.

• Read .vibeflow/index.md or project config files (package.json, pyproject.toml, Cargo.toml, etc.) to detect the test runner.
• Always include the detected test command (e.g., npm test, pytest, cargo test, go test ./...).
• If the spec lists specific test commands, include those too.
• If no test runner is detected, write: "No test runner detected. Add manual tests to validate."
• Format example:

  ## How to validate
  1. Run tests: `npm test`
  2. Verify manually: [description]
  

8. Docs to Update

Which docs need changes after implementation.

If you cannot verify a file path exists, flag it with: <!-- TODO: verify this path -->

Save the prompt pack to: .vibeflow/prompt-packs/<feature-slug>.md Create the .vibeflow/prompt-packs/ directory if it doesn't exist.

After saving, suggest: "Prompt pack saved. Hand it off to the coding agent. After implementation, run /vibeflow:audit with the spec to verify."

---

Maintenance

If this command is modified, update MANUAL.md to reflect the changes.