From what-do-you-know
Deep-dive into a topic using your arsenal evidence, with persona-driven probing questions and concept mapping
How this skill is triggered — by the user, by Claude, or both
Slash command
/what-do-you-know:explore-wdykThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
<references>
<action>Collect all matching entries as context for the session.</action>
Load Session History
bash /scripts/search-sessions.sh --topic --recent 5
Use previous session data to avoid re-asking the same probing questions and to track progress on this topic.
Select Persona
Use the specified persona.
Use the `default_persona` from config.
Load the selected persona from `/.things/shared/roles/.md`.
Adopt the persona's voice, evaluation weights, and follow-up patterns -- but adapt them for learning rather than interview coaching. The persona is probing to understand and deepen knowledge, not to evaluate interview readiness.
Select Depth
Use the specified depth.
Use `default_depth` from config.
- exploratory -- Broad overview: "What do you know about X? Walk me through the landscape." Cover breadth, identify strong and weak areas.
- focused -- Targeted deep-dive: "Let's go deep on X. Explain how it works under the hood." Probe internals, edge cases, tradeoffs.
- deep -- Intensive probing: "You've worked with X. Let's stress-test your mental model." Challenge assumptions, explore failure modes, demand precision.
Begin Probing Dialogue
Use the persona's voice to begin exploring the topic. Ground questions in the user's actual experience from their arsenal.
Question patterns:
- Experience-grounded: "You made an architecture decision about [X] in [project from arsenal]. Walk me through your reasoning."
- Scale challenge: "What would break first if you scaled this 10x?"
- Comparison: "How does your approach compare to [standard approach]?"
- Edge cases: "What happens when [edge case]? How would you handle it?"
- Teaching test: "If you had to explain [concept] to a junior engineer, where would you start?"
- Connection probing: "How does [concept A] relate to [concept B from their experience]?"
- Gap detection: "What's the difference between [similar concept A] and [similar concept B]?"
<constraint>
Adapt the probing based on depth level:
- exploratory: 4-6 questions, broad coverage, identify landscape
- focused: 6-8 questions, targeted depth in specific areas
- deep: 8-12 questions, intensive probing, challenge mental models
</constraint>
Probe Deeper Where Strong, Widen Where Thin
As the user responds, adapt the probing strategy.
Push deeper: "What's happening at the layer below that?" or "What are the tradeoffs of that approach vs. alternatives?"
Widen: "Let's set that aside for now. What about [related concept]?" or "That's a gap worth noting. Let me ask about something adjacent."
Connect: "You mentioned [entry]. How does that experience inform your understanding of the general principle?"
Use the `session_length` config to pace the conversation:
- short: ~15 min, 4-5 exchanges
- medium: ~30 min, 7-8 exchanges
- deep: ~60 min, 12+ exchanges
Score on Learning Dimensions
After the dialogue, score the user on five dimensions (from `references/exploration-framework.md`):
1. Depth (1-5) -- How far below the surface can they explain? Internals vs. buzzwords
2. Accuracy (1-5) -- Are their mental models technically correct?
3. Connections (1-5) -- Can they relate this to adjacent concepts and their own experiences?
4. Application (1-5) -- Can they apply the knowledge to new situations?
5. Articulation (1-5) -- Can they explain it clearly to someone else?
Produce Concept Map
Classify concepts discussed into strong, partial, and gap categories.
- Strong -- demonstrated depth, accuracy, and ability to teach
- Partial -- understands the basics but gaps in depth or precision
- Gap -- couldn't explain or had significant misconceptions
<template name="concept-map">
<output>
Concept Map: [Topic]
Strong:
- [concept] -- [evidence of strength]
Partial:
- [concept] -- [what's missing]
Gap:
- [concept] -- [what to learn]
</output>
</template>
Deliver Feedback
Use the persona's voice to deliver feedback.
<template name="persona-assessment">
<output>
[Persona Name]'s Assessment
| Dimension | Score | Notes |
|-----------|-------|-------|
| Depth | x/5 | brief note |
| Accuracy | x/5 | brief note |
| Connections | x/5 | brief note |
| Application | x/5 | brief note |
| Articulation | x/5 | brief note |
Strongest area: what they demonstrated well
Biggest opportunity: most impactful area to deepen
From your arsenal: connections to logged experiences that could ground further learning
</output>
</template>
Log the Session
<schema name="session-frontmatter">
```markdown
---
type: explore
date: <date>
topic: <topic>
persona: <persona>
depth: <exploratory|focused|deep>
scores:
depth: <1-5>
accuracy: <1-5>
connections: <1-5>
application: <1-5>
articulation: <1-5>
overall: <average>
strong_concepts:
- <concept>
partial_concepts:
- <concept>
gap_concepts:
- <concept>
arsenal_references:
- file: "<log filename>"
relevance: "<how it was used>"
---
```
</schema>
<section name="body">
```markdown
# Explore Session: <topic> -- <date>
## Topic
<topic>
## Concept Map
<strong / partial / gap breakdown>
## Key Exchanges
<summary of most revealing Q&A moments>
## Feedback
<summary of feedback given>
## Recommended Next Steps
<what to explore, quiz, or bridge next>
```
</section>
</write>
Update Knowledge Map and Progress
Read and update `/.things/what-do-you-know/knowledge-map.json`:
- Add or update topic entries under strong / building / gap / blind_spot based on scores
- Note specific concepts in each category
Read and update `/.things/what-do-you-know/progress.json`:
- Update dimension averages with new scores
- Update topic breakdown
- Add to sessions list
Handle Git Workflow
Before committing, pull latest changes from the remote (if one exists) to avoid conflicts.
git -C /.things pull --rebase 2>/dev/null || true
Based on the `git_workflow` setting in `config.json`:
Would you like to commit and push this explore session?
Yes
No
Automatically `git add` the session file, knowledge-map.json, and progress.json, `git commit -m "explore: "`, and `git push`.
Tell the user the session has been saved and they can commit when ready.
Suggest Next Steps
<ask-user-question>
<question>What next?</question>
<option>Explore another topic</option>
<option>Quiz this topic</option>
<option>Bridge a gap from this session</option>
<option>Done for now</option>
</ask-user-question>
</completion-message>
npx claudepluginhub brennacodes/brenna-plugs --plugin what-do-you-knowDelivers structured multi-session tutoring for technical topics with Socratic diagnostics, knowledge graphs for prerequisites, agendas, teaching, quizzes, and progress tracking.
Formative assessment of a user's understanding of a topic or code. Returns a discrete Bloom/SOLO level, concrete gaps, and a suggested next skill.
Guides structured learning of new topics by assessing knowledge, designing paths, adapting difficulty, and planning review sessions. Use when overwhelmed by docs or needing spaced repetition.