Research Gather Skill

Purpose: Systematically assemble all source material needed for an article.

When to Use

At the RESEARCH stage of the blog pipeline, after IDEA is defined and before WALK.

Source Types

Source What to Gather How

GitHub history Commits, PRs, issues relevant to story git log , GitHub API

Session transcripts Past Claude conversations on topic Search transcripts collection

Code examples Actual code that demonstrates points Extract from repos

Screenshots UI states, terminal output, diagrams Capture or find existing

Dates & timeline When things happened Git history, file dates

External references Docs, articles, tools mentioned Web search, bookmarks

Process

0. Load Author Context

Before any research, read from eterne (private vault):

~/src/eterne/vault/03-resources/blog-context/ ├── author.md # Quick reference (always load) ├── timeline/overview.md # For chronological articles ├── people/family.md # If relationships mentioned └── themes/ ├── spiritual.md # If practice/philosophy relevant └── career.md # If professional history relevant

Why eterne? Biographical context is private (names, health history, location).

This provides:

• Verified biographical facts (prevents hallucination)

• Current project metrics and scale

• Key claims with evidence status

• Family and relationship context

• Spiritual/philosophical background

• Career arc (including illness era context)

1. Define Research Questions

Based on the one-sentence teaching from IDEA stage:

• What claims will this article make?

• What evidence supports each claim?

• What concrete examples illustrate each point?

2. Gather by Source Type

GitHub History:

Find commits related to topic

git log --oneline --grep="[keyword]" git log --oneline --since="2024-01-01" --until="2024-06-01" -- path/to/relevant/

Find PRs

gh pr list --state all --search "[keyword]"

Session Transcripts:

• Search packages/blog/src/content/transcripts/

• Search conversation history for relevant topics

• Note session dates and key exchanges

Code Examples:

• Extract minimal, self-contained examples

• Verify they still run

• Note file paths and commit hashes for attribution

Timeline:

• Build chronological sequence of events

• Note key dates, durations, milestones

3. Organize Material

Create research file at pipeline/active/NNN-slug/02-research.md :

Research: [Article Title]

Claims & Evidence

Claim 1: [Statement]

• Evidence: [source, quote, data]
• Example: [code/screenshot/transcript excerpt]

Claim 2: [Statement]

...

Timeline

Code Examples

Example 1: [Description]

[code]

Source: [file path @ commit]

Open Questions

- [Things to explore during WALK stage]

- [Gaps in evidence]

Raw Material

[Links, excerpts, screenshots to process]

---

## Quality Gate

**Gate question:** Do you have concrete examples for every claim?

**Checklist:**
- [ ] Author context loaded and referenced
- [ ] Every claim has supporting evidence
- [ ] Timeline is documented with sources
- [ ] Code examples are extracted and verified
- [ ] Screenshots captured if needed
- [ ] Gaps identified for WALK stage exploration

---

## Anti-Patterns

| Anti-Pattern | Problem | Fix |
|--------------|---------|-----|
| Vague notes | Can't verify later | Cite specific sources |
| Too much material | Overwhelms outline | Focus on claims, not completeness |
| Unverified examples | May not work | Run code, check links |
| Missing dates | Timeline unclear | Always note when things happened |

---

*Research serves the story. Precision now prevents problems later.*