Prompt &
Agent Knowledge
For those who want to build the weapons themselves. Battle-tested prompt frameworks and the agent design playbook.
The question your client is waiting for.
Each template is the output of real client work. Adapt the structure; replace the variables; run it.
Meeting Minutes → Action Items
Extract action items from the following meeting notes.
For each action, include: "Owner", "Deadline", "Priority (H/M/L)".
If unknown, mark as "TBD".
[MEETING NOTES]
{Paste meeting notes here}
Client Brief Summary
Structure the following client brief into:
1. Background / Challenge (3 lines max)
2. Objective (1 sentence)
3. Target Audience
4. KPIs
5. Timeline
6. Budget Range
7. Constraints / Notes
[BRIEF]
{Paste brief here}
Estimate Draft
Create a cost estimate structure for the following scope.
Include: "Line item", "Man-days", "Approx. unit cost", "Subtotal".
Format as a table.
Note: Actual rates will be adjusted to internal standards.
[PROJECT SCOPE]
{Describe scope here}
Competitive Landscape Analysis
Analyze the competitive landscape for {brand name}.
1. 3-5 direct competitors and their positioning
2. Recent moves (last 6 months) per competitor
3. Differentiation comparison table
4. White space hypothesis
Market: {Thailand / Indonesia / ASEAN}
Category: {category name}
STP Hypothesis Builder
Based on the following information, create 3 STP
(Segmentation / Targeting / Positioning) hypotheses.
For each pattern:
- Segmentation axis & rationale
- Target segment definition & estimated size
- Positioning statement (1 sentence)
- 3 Reasons to Believe
[INPUT]
{Paste brief, market data, research, etc.}
Sei-katsu-sha Insight Discovery
Act as a Sei-katsu-sha matching this profile: {target description}.
Answer the following honestly, as emotional truth, not surface-level:
- If {brand/category} disappeared tomorrow, what would feel missing?
- When was the last time you felt "this is it" in this category?
- What do you secretly want from this category but never say?
Contradictions are welcome. Humans are contradictory beings.
Copy Variation Generator
Generate 10 copy variations for social media
from the following core message.
Conditions:
- Platform: {Instagram / TikTok / LINE}
- Tone: {upbeat / serious / provocative / empathetic}
- Length: {under 30 chars / under 60 / under 120}
- Hashtags: 5 per variation
- Emoji: {yes / no}
[CORE MESSAGE]
{Paste original copy here}
Creative Brief Expansion
From the following creative brief, propose 3 distinct
creative directions.
For each direction:
- Concept word (2-3 words)
- Visual tone & manner
- Key copy (1 sentence)
- Risks and opportunities of this approach
[BRIEF]
{Paste creative brief here}
Campaign Report Narrative
Create a monthly report narrative from the following data.
Structure:
1. Executive Summary (3 lines)
2. Key KPI performance (MoM & vs. target)
3. Top 3 Insights (data-driven findings)
4. Issues & improvement recommendations
5. Next month action plan
[DATA]
{Paste KPI data here}
Media Plan Rationale
Write a client-facing rationale for the following media plan.
For each medium:
- Why this medium (target touchpoint rationale)
- Expected effect
- Budget allocation rationale
[MEDIA PLAN]
{Paste media plan summary here}
Getting Started with Claude Code
Use AI directly from the terminal. File operations, code generation, and agent building — a practical guide.
Core Concepts
Agent-Based AI
Not chat — execution. AI that reads and writes files, runs code, and autonomously completes multi-step tasks.
Skills & MCP
Custom skills (/deck, /translate, etc.) automate workflows. MCP extends tool connectivity.
Memory & Context
CLAUDE.md defines project rules. Memory files maintain learning across sessions.
Usage Patterns
Meeting → Auto Deck Generation
Input meeting transcript → analyze → auto-generate purpose-specific decks (Strategic / Learning / Creative / EN) in one pass.
PPTX Translation (Format-Preserving)
Run-level translation preserving layout, fonts, and colors. JP → EN / TH / ID / VN.
Data Analysis Pipeline
/analyze → /strategy → /deck → /translate → /review. End-to-end AI pipeline.
Agent Architecture Framework
The foundational framework for designing your own AI agents.
Define Task
What to automate. What goes in, what comes out. Identify the highest time-consuming process.
Design Knowledge
What should the agent know? Past cases, brand guidelines, industry data. Inject external knowledge via RAG.
Design Prompts
System prompt, few-shot examples, output format definition. Persona settings. Error handling.
Test & Iterate
Test with 10 cases. Quality checklist. Feedback loop. Iterative improvement: v1 → v2 → v3.