Skip to content

po4yka/llm-wiki-skills / skills/llm-wiki-interview

llm-wiki-interview

Verified

Extract tacit knowledge into an LLM-Wiki through an agent-led interview. Use when the wiki has open questions, weak areas, onboarding gaps, undocumented decisions, or the user wants voice/text answers turned into draft pages without pretending they are sourced facts.

LLM-Wiki Interview

Goal

Use targeted questions to capture a person's tacit knowledge into draft, reviewable wiki pages.

When to use

  • The wiki has open questions, sparse concept pages, or pages marked review_required.
  • Onboarding gaps keep producing the same repeated questions from new contributors.
  • A decision page is missing its rationale, history, or trade-offs.
  • The user hands over a voice-note or transcript that should become draft wiki pages.
  • A lint report or contradiction check surfaces a gap only a person can fill in.

Inputs

  • wiki/index.md, wiki/log.md, open questions and lint reports.
  • Target domain or project.
  • Interview mode: plan questions, conduct interview, or file answers.
  • Optional transcript or voice-note text.

Procedure

1. Find weak areas

Look for:

  • open questions;
  • sparse concept pages;
  • onboarding gaps;
  • contradictions;
  • pages marked review_required;
  • repeated user questions;
  • decision pages missing rationale.

Search wiki/index.md and wiki/log.md for prior interview answers before drafting new questions, so the interview does not re-ask what is already captured.

2. Prepare focused questions

Ask 5-10 questions at a time. Prefer questions that extract:

  • constraints;
  • history;
  • rationale;
  • exceptions;
  • gotchas;
  • examples;
  • terminology;
  • decision trade-offs.

3. Capture answers as draft knowledge

Store interview-derived knowledge as:

  • type: query for Q&A;
  • type: synthesis for human interpretation;
  • type: concept or entity only when the answer clearly belongs there.

Mark:

status: draft
review_required: true
claim_mix:
  extracted: 0.0
  inferred: 0.0
  ambiguous: 0.0

Use interview or human-memory source tags. Do not treat interview answers as external facts.

4. Separate human synthesis

Preserve answers under human-owned sections when they represent the person's interpretation.

5. Update navigation

Add important pages to wiki/index.md and append to wiki/log.md if writing files.

Output

## Interview plan

## Questions

## Captured answers

## Draft pages created or proposed

## Claims needing external sources

## Follow-up questions

Safety gates

  • Do not present interview answers as source-verified facts.
  • Do not pressure the user with too many questions at once.
  • Do not overwrite existing human synthesis.
  • Do not publish private answers without approval.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall