Showcase Export Mode

Ensures all skill/agent orchestration is VISIBLE in the session transcript when you run the built-in /export command.

Note: This skill defines a PROTOCOL for what Claude should narrate during a session. It works WITH the built-in /export command, not as a replacement.

Quick Start

bash

# At session START
"Build my-project --showcase"

# When done, use built-in export
/export                    # Export to clipboard or file

For OLD sessions, use the session-reconstruct skill instead.

The Problem

code

Default behavior:
- Skills execute silently (instructions inject invisibly)
- Agents work in background (subagent reasoning hidden)
- Compound learning runs as bash commands (semantic meaning lost)
- Decision rationale is implicit (tradeoffs not explained)
- /export only captures visible conversation
- YC doesn't see your orchestration mastery

The Four Structural Gaps

Gap	What's Hidden	Fix Required
Skill Logic	Skill instructions inject into context invisibly	Explain what the skill instructed you to do
Subagent Internals	Task tool returns only final result	Narrate the agent's internal process
Compound Learning	DB operations are opaque bash commands	Explain the semantic meaning of patterns
Decision Rationale	Choices made without explaining tradeoffs	State every significant tradeoff explicitly

The Solution

Every skill, agent, compound update, and decision must be narrated for the transcript.

This is not optional verbosity—it's structural visibility that makes orchestration mastery observable in /export.

Mandatory Orchestration Protocol

When --showcase is enabled, you MUST follow ALL of these protocols:

1. Announce Skill Activation (with Logic Surfacing)

When a skill loads, its instructions are invisible to the transcript. You MUST explain what the skill is telling you to do:

markdown

## 🔧 Skill Activated: [skill-name]

**Purpose:** [what this skill does]
**Triggered by:** "[user's words that triggered it]"

**This skill instructs me to:**
1. [First instruction from the skill]
2. [Second instruction from the skill]
3. [Third instruction from the skill]

**I will now execute these instructions...**

2. Show Skill Execution (with Reasoning)

markdown

### Executing: [skill-name]

**Step 1:** [what we're doing]
→ Reasoning: [why this step matters]
→ Result: [outcome]

**Step 2:** [what we're doing]
→ Alternative considered: [what else could have been done]
→ Why rejected: [reason]
→ Result: [outcome]

3. Summarize Skill Output (with Tradeoff Documentation)

markdown

### ✅ Skill Complete: [skill-name]

**Deliverables:**
- [output 1]
- [output 2]

**Key decisions made (with tradeoffs):**
| Decision | Alternative | Why Chosen |
|----------|-------------|------------|
| [Choice A] | [Alternative B] | [Specific reason] |

4. Announce Agent Spawning (with Internal Process Narration)

When spawning agents via Task tool, only the final result returns. You MUST reconstruct and narrate what the agent did:

markdown

## 🤖 Spawning Agent: [agent-name]

**Mission:** [what this agent will do]
**Tools available:** [list of tools]
**Why this agent vs. doing it directly:** [reason for delegation]

---

### Agent Complete: [agent-name]

**Final Result:** [the returned summary]

**Reconstructed Internal Process:**
- **Tool calls made:** ~[estimate] ([list tools likely used])
- **Key reasoning steps:**
  1. [Inferred reasoning step 1]
  2. [Inferred reasoning step 2]

**Decision points the agent navigated:**
| Decision | Agent's Choice | Likely Reason |
|----------|----------------|---------------|
| [Decision 1] | [Choice] | [Reason] |

5. Document Decision Points

markdown

### 📋 Decision Point: [Brief Description]

**Options Considered:**
| Option | Pros | Cons |
|--------|------|------|
| [A] | [advantages] | [disadvantages] |
| [B] | [advantages] | [disadvantages] |

**Decision:** [What was chosen]
**Rationale:** [Why this option won]

6. Show Compound Learning (Semantic Meaning)

markdown

### 🔄 Compound Learning Update

**Pattern Extracted:**
> "[Natural language description of what was learned]"

**Evidence:**
1. [What demonstrated this]

**How this changes future behavior:**
Before: [old behavior]
After: [new behavior]

7. Checkpoint Summaries

markdown

───────────────────────────────────────────────────────
### 📊 Checkpoint: After Phase [N]
───────────────────────────────────────────────────────

**Skills used:** [count] ([list])
**Agents spawned:** [count] ([list])
**Key decisions:** [list with rationale]

**Proceeding to Phase [N+1]...**

Export Checklist

Before running /export, verify:

• Every skill activation explains what the skill instructed
• Every agent spawn includes reconstructed internal process
• Compound learning includes semantic meaning
• Every significant decision shows tradeoffs
• Phase transitions clearly marked
• Final summary included

Complete Workflow

bash

# 1. START SESSION with showcase mode
"Build my-project --showcase"

# 2. WORK normally - Claude narrates orchestration per this protocol

# 3. EXPORT using built-in command
/export session.md

# 4. (Optional) RECONSTRUCT if gaps exist
"Fill in any orchestration gaps --reconstruct"

Installation

bash

npx skills add sunnypatneedi/skills