Plan Preview Workflow

Purpose

Generate concise, stakeholder-ready summaries from detailed implementation plans. Enables developers, tech leads, and project owners to understand and approve planned changes without reviewing full technical details.

Input Sources

Accept plan from one of:

{TYPE}-{NUMBER}@{HASH}

PLAN-0295@a1b2c3

.agent/issues/references/PLAN-0295@a1b2c3-plan.md

Resolution order:

1. If issue ID provided → resolve to

   .agent/issues/references/{id}-plan.md

2. If path provided → read directly
3. If neither → check if plan exists in current conversation context

Procedure

Step 1: Resolve Input

IF issue_id provided:
    path = .agent/issues/references/{issue_id}-plan.md
    IF NOT exists(path):
        ERROR "Plan file not found for issue {issue_id}"
ELSE IF file_path provided:
    path = file_path
ELSE:
    Scan conversation context for plan content

Step 2: Language Selection

Ask user (one question):

"What language should the summary be in? (default: English)"

Common choices:

• English (default)
• Norwegian (Norsk)
• Other (specify)

Wait for response before proceeding.

Step 3: Confidence Level Selection

Ask user (one question):

"What confidence level is this plan? This affects detail level in the summary:

• LOW — More details, explicit changes, method signatures
• NORMAL — Balanced overview (default)
• HIGH — Sparse, broad outlines only"

Wait for response before proceeding.

Step 4: Extract Plan Elements

Parse the implementation plan and extract elements based on confidence level:

Detail Level by Confidence

Step 5: Flow Diagram Decision

Confidence-aware diagram rules:

Standard criteria (NORMAL confidence):

• Plan has branching logic (if/else paths)
• Plan has multi-step sequences with dependencies
• Plan involves state transitions
• Plan has complex component interactions

Step 6: Generate Summary

Structure varies by confidence level:

LOW Confidence Template

# Plan Summary: {objective_short}

**Issue**: {issue_id}
**Generated**: {date}
**Language**: {language}
**Confidence**: LOW (detailed preview)

## Objective

{2-3 sentences with full context and background}

## Approach

{5-7 sentences explaining the implementation strategy}

**Edge Cases Considered:**
- {edge case 1}
- {edge case 2}

## Files Changed

| File | Change Type | Lines (est.) | Details |
|------|-------------|--------------|---------|
| `path/to/file.ts` | New | ~150 | {detailed description of what file contains} |
| `path/to/existing.ts` | Modify | +30/-10 | {specific changes being made} |

## New Components

### {Category}

| Name | Signature | Purpose |
|------|-----------|---------|
| `methodName` | `(param: Type) → ReturnType` | {detailed description} |

## Dependencies

| Package | Version | Reason |
|---------|---------|--------|
| `package-name` | ^1.2.3 | {why this dependency is needed} |

## Flow Overview

{Mermaid diagram - always included for LOW confidence}

## Risks & Considerations

| Risk | Impact | Mitigation |
|------|--------|------------|
| {risk description} | {HIGH/MEDIUM/LOW} | {how to address} |

---
*Detailed preview for LOW confidence plan — verify assumptions before implementation.*

NORMAL Confidence Template

(Use existing Step 5 template from original skill)

HIGH Confidence Template

# Plan Summary: {objective_short}

**Issue**: {issue_id}
**Generated**: {date}
**Confidence**: HIGH (executive summary)

## Objective

{1 sentence}

## Approach

{1 sentence}

## Scope

- **Files**: {count} files ({new_count} new, {modify_count} modified)
- **Components**: {component_summary, e.g., "2 endpoints, 1 service, 3 tests"}
- **Dependencies**: {Yes/No new dependencies}

## Key Risks

{Only if critical risks exist, otherwise omit section}

---
*High-confidence plan — minimal review needed.*

Step 7: Present & Offer Options

Display the generated summary, then ask:

"Summary generated. What would you like to do?

1. Open in editor — Opens as new file in VS Code
2. Save to .agent/docs/ — Saves as

   {issue_id}-preview.md

3. Done — Keep in chat only"

Output Templates by Language

English (default)

Use templates from Step 6 based on confidence level.

Norwegian (Norsk)

NORMAL Confidence (Norsk)

# Plansammendrag: {objective_short}

**Sak**: {issue_id}
**Generert**: {date}
**Språk**: Norsk
**Konfidensnivå**: NORMAL

## Mål

{1-2 setninger om hovedmålet}

## Tilnærming

{2-3 setninger om hvordan planen oppnår målet}

## Endrede filer

| Fil | Endringstype | Formål |
|-----|--------------|--------|

## Nye komponenter

### {Kategori: Metoder/Endepunkter/Validatorer/Kommandoer}

| Navn | Formål |
|------|--------|

## Avhengigheter

## Flyt

## Risiko og vurderinger

---

*Dette sammendraget ble generert fra implementeringsplanen for interessentgjennomgang.*

HIGH Confidence (Norsk)

# Plansammendrag: {objective_short}

**Sak**: {issue_id}
**Generert**: {date}
**Konfidensnivå**: HØY (sammendrag)

## Mål

{1 setning}

## Tilnærming

{1 setning}

## Omfang

- **Filer**: {antall} filer ({nye} nye, {endrede} endrede)
- **Komponenter**: {komponentsammendrag}
- **Avhengigheter**: {Ja/Nei nye avhengigheter}

---
*Høy-konfidensplan — minimal gjennomgang nødvendig.*

Completion Criteria

• Plan source resolved (issue ID, path, or context)
• Language confirmed with user
• Confidence level confirmed with user
• Summary generated with appropriate detail level
• Mermaid diagram included according to confidence rules
• Output presented to user
• User chose action (editor/save/done)

Anti-patterns (avoid)

• ❌ Generating output before asking for language preference
• ❌ Generating output before asking for confidence level
• ❌ Including Mermaid diagrams for HIGH confidence plans
• ❌ Providing sparse details for LOW confidence plans
• ❌ Copying implementation details verbatim (defeats purpose)
• ❌ Assuming English without asking
• ❌ Assuming NORMAL confidence without asking
• ❌ Saving file without user consent

Examples

Example 1: HIGH Confidence (Brief Overview)

User: "Preview plan for FEAT-0042@abc123"

Agent:

1. Resolves

   .agent/issues/references/FEAT-0042@abc123-plan.md

2. Asks: "What language?" → User: "English"
3. Asks: "What confidence level?" → User: "High"
4. Generates brief summary (no diagram):

# Plan Summary: Add User Settings API

**Issue**: FEAT-0042@abc123
**Confidence**: HIGH (overview)

## Objective
Add CRUD endpoints for user settings.

## Approach
RESTful API with validation middleware.

## Scope
- **Files**: 4 files (2 new, 2 modified)
- **Components**: 1 service, 4 endpoints
- **Dependencies**: None new

Example 2: NORMAL Confidence (Standard)

User: (plan in context) "Create preview in Norwegian"

Agent:

1. Detects plan in context
2. Language already specified: Norwegian
3. Asks: "What confidence level?" → User: "Normal"
4. Generates standard summary with optional diagram (if workflow has branches)

Example 3: LOW Confidence (Detailed Review)

User: "Preview this complex migration plan, I need detailed review"

Agent:

1. Detects "detailed review" signals LOW confidence
2. Asks: "What language?" → User: "English"
3. Confirms: "LOW confidence for detailed review. Correct?"
4. Generates comprehensive summary with detailed tables, diagrams, and risk matrix

Example 4: Diagram Decision by Confidence