seminar

Seminar: Cross-Examination of Curated Knowledge

Transform curated research into a defensible position through structured interrogation.

Read ${CLAUDE_PLUGIN_ROOT}/skills/seminar/references/examination-techniques.md for the four techniques. NotebookLM scripts: Use ${CLAUDE_PLUGIN_ROOT}/scripts/notebook-ask.sh for all notebook queries. Reference: ${CLAUDE_PLUGIN_ROOT}/skills/literary-review/references/notebooklm-cli.md.

---

Input

Required:

• Notebook ID — a NotebookLM notebook with curated sources
• Leading questions — specific questions to drive the examination
• User context — constraints, preferences, existing knowledge

Optional:

• Workshop findings — 03-workshop.md if running in pipeline
• Engagement directory — for writing output

---

Step 1 — Frame the Examination

From the leading questions and context, identify:

• What decision needs to be made? (not "learn about X" but "decide between A and B")
• What assumptions are we carrying? (list them explicitly to challenge)
• What would change our mind? (define the falsification criteria)

---

Step 2 — Apply Examination Techniques

Use all four techniques from examination-techniques.md. For each, query the notebook with structured questions.

Pattern Spotting

Ask the notebook: "What patterns appear across multiple sources on ? Which patterns are supported by the most evidence?"

Paradox Hunting

Ask the notebook: "What seems true about based on the sources but might not be? Where do sources contradict each other or contradict common wisdom?"

Naming the Unnamed

Ask the notebook: "What recurring concepts in the sources don't have a standard name? What behaviors or patterns are described but not labeled?"

Give discovered patterns names. These become the engagement's vocabulary.

Contrast Creation

Ask the notebook: "Compare vs based on the sources. What does each gain? What does each lose? When would you choose one over the other?"

---

Step 3 — Challenge Assumptions

For each assumption identified in Step 1:

${CLAUDE_PLUGIN_ROOT}/scripts/notebook-ask.sh NOTEBOOK_ID \
  "What evidence supports the assumption that <assumption>? What evidence contradicts it?"

Record: assumption, supporting evidence, contradicting evidence, revised confidence.

---

Step 4 — Synthesize Output

Write 04-seminar.md:

# Seminar: <topic>

## Decision Frame
<What decision does this inform?>

## Named Concepts
<New vocabulary — patterns we identified and named>
- **<Name>**: <definition and evidence>
- **<Name>**: <definition and evidence>

## Decision Table

| Option | Strengths | Weaknesses | When to Use | Risk |
|--------|-----------|------------|-------------|------|
| ... | ... | ... | ... | ... |

## Ranked Hypotheses

1. **<Hypothesis>** — Confidence: <high/medium/low>
   Evidence for: <summary>
   Evidence against: <summary>

2. **<Hypothesis>** — Confidence: <high/medium/low>
   ...

## Assumptions Challenged
| Assumption | Supported? | Key Evidence |
|------------|-----------|--------------|
| ... | ... | ... |

## Open Questions
<What we still don't know and can't answer from the sources>

---

Standalone Mode

When used outside of research-lab:run:

1. Ask the user for the notebook ID, leading questions, and context
2. Run Steps 1-4
3. Present the output and offer:
   • "Ready to design a methodology based on these findings?"
   • "Want to run an experiment testing the top hypothesis?"
   • "Archive to vault?"

Vault archival (standalone)

VAULT_ROOT="$HOME/Vaults/${OBSIDIAN_VAULT_NAME:-Neurons}"
TOPIC_SLUG="<kebab-case-topic>"
DEST="Research/$TOPIC_SLUG/$(date +%Y-%m-%d)-seminar.md"
mkdir -p "$VAULT_ROOT/$(dirname "$DEST")"
cp "04-seminar.md" "$VAULT_ROOT/$DEST"