Note Taking

Overview

Maintain an audit trail of discussions, decisions, and progress for each blog post or newsletter. The discussion.md file serves two purposes:

For Claude: Quick context to avoid repetition and maintain holistic view
For User: Audit trail showing what Claude did, what decisions were made, and why

Core principle: Track attribution (what Claude did vs. what user decided) and outcomes (did it work?).

Announce at start: "I'm recording this discussion in discussion.md."

Success Criteria

The discussion record is complete when:

Decision attribution is clear - Shows what Claude contributed vs. what user decided
Cause and effect tracked - "Claude did X → User decided Y because Z"
Outcomes documented - Did the decision work? What happened?
Quick lookup enabled - Can find key decisions without reading full narrative
Audit trail exists - Can trace back why any decision was made

File Structure

discussion.md follows this structure:

# Discussion Notes: [Project Name]

## AUDIT TRAIL: Key Decision Points

[Decision-by-decision record - see template below]

## WHAT CLAUDE DID (Contributions)

### Research
- [Bullet list of research performed]

### Writing
- [Bullet list of drafts, rewrites, sections created]

### Quality Control
- [Bullet list of audits, reviews, validations]

### Technical
- [Bullet list of scripts, tools, automation]

## WHAT WORKED / DIDN'T WORK

### Worked Well ✅
| What Claude Did | User Decision | Outcome |
|-----------------|---------------|---------|
| [action] | [decision] | [result] |

### Didn't Work ❌
| What Claude Did | Problem | Lesson |
|-----------------|---------|--------|
| [action] | [what failed] | [what to avoid] |

## SESSIONS (Chronological Detail)

[Full session-by-session narrative for context]

The Process

Step 1: Determine the Post Folder

If working on existing post: locate /posts/{post-short-title}/
If creating new post: folder created during pitch generation
If unsure: ask user which post this relates to

Step 2: Read Existing discussion.md

CRITICAL: Always read existing discussion.md before starting work.

When reading, extract:

What decisions have been made (don't re-debate)
What content is already covered (don't repeat)
What approaches failed (don't retry)
What Claude already contributed (build on it)

Step 3: Record the Session

Add session detail at bottom, then update top sections.

For Each Session: Add to Bottom

## Session X: [Short Title] (YYYY-MM-DD)

### Context
[Why this session happened, what triggered it]

### What Claude Did
- Research: [what research]
- Analysis: [what analysis]
- Writing: [what drafted/revised]
- Tools: [what tools/scripts created]

### User Decisions
- Decision 1: [what was decided and why]
- Decision 2: [what was decided and why]

### Outcomes
- ✅ What worked: [successes]
- ❌ What didn't: [failures]
- Files created/modified: [list]

### Next Steps
[What comes next]

For Major Decisions: Add to AUDIT TRAIL

Use this template for significant decisions:

### Decision: [Decision Title] (YYYY-MM-DD, Session X)

**Claude's Analysis:**
- [What research/analysis Claude performed]
- [What data/insights Claude provided]
- [What options Claude presented]

**Claude's Recommendation:**
[What Claude suggested and why]

**User Decision:**
[What the user actually decided]

**Rationale:**
[Why the user made this choice]

**Outcome:**
✅/❌ [What happened as a result]

Step 4: Update Top Sections

After adding session details, update:

AUDIT TRAIL - Add major decisions with attribution
WHAT CLAUDE DID - Add to contribution categories
WHAT WORKED / DIDN'T WORK - Update outcome tables

Step 5: Save the Record

Append to /posts/{post-short-title}/discussion.md
Use Edit tool to update top sections
Don't replace existing content - build on it

When to Use This Skill

Always use after:

Brainstorming sessions
Generating or revising pitches
Significant research or feedback
Quality control reviews
Publishing to Ghost
Major milestones in writing process

Also use when:

User explicitly requests it
Making major decisions
Something fails or succeeds notably
End of a series or project

Recording Decisions: Best Practices

Good decision record:

### Decision: Use "150+ hours" not "600 hours" (2025-11-03)

**Claude's Analysis:**
- Git analysis: 79 commits across 36 days
- Found error: content is automated, not manually curated
- Revised estimate: 150-200 hours based on commit patterns

**Claude's Recommendation:**
Use "over 150 hours across several months"

**User Decision:**
Approved. Changed throughout Part 1.

**Rationale:**
More honest (admits uncertainty), shows AI's impact, still significant.

**Outcome:**
✅ More credible than 600-hour claim. Readers responded well.

Bad decision record:

We decided to use 150 hours instead of 600.

(Missing: what Claude did, why decision was made, what happened)

What Makes Good Audit Trail

Capture:

✅ What Claude analyzed/researched
✅ What Claude recommended
✅ What user decided (might differ from recommendation!)
✅ Why user made that choice
✅ What happened as result

Avoid:

❌ Narrative prose without attribution
❌ Decisions without rationale
❌ Recommendations without alternatives considered
❌ Outcomes without assessment (worked or didn't?)

For Series Work

Additional tracking needed:

What's been covered - Track to avoid repetition

## WHAT'S BEEN COVERED
### Part 1 established:
- Domain renewal decision
- Framework: assess, identify, clarify
- Zero users truth

### Part 2 established:
- Three-layer architecture
- Cost breakdown
- Time savings

Series state - Track what's done/pending

## SERIES STATE
- Part 1: PUBLISHED (Nov 7)
- Part 2a: PUBLISHED (Nov 14)
- Part 2b: POSTPONED
- Part 3: PUBLISHED (Nov 22)

Cross-series lessons - What worked across all parts

## SERIES LESSONS
- ✅ Research before writing (validated decisions)
- ✅ Multiple reviewer perspectives
- ❌ Treated parts as standalone (repeated context)
- ❌ Didn't read discussion.md first

Files in This Skill

SKILL.md (this file) - Main skill instructions
decision-template.md - Template for recording decisions
session-template.md - Template for recording sessions

Remember

For Claude:

Read discussion.md FIRST before any work
Extract what's covered to avoid repetition
Check what failed to avoid retrying

For User (audit trail):

Show what Claude did that led to decisions
Attribute clearly (Claude's work vs. User's choice)
Track outcomes (worked or didn't?)
Enable tracing: "Why did we decide X?" → full path visible

For both:

Decisions need context and rationale
Outcomes need assessment
Lessons learned feed improvement

The goal: Create an audit trail showing what Claude contributed, what decisions resulted, and whether they worked.