Use this skill whenever the user wants to discuss, explore, or ask questions about a topic using the local knowledge base as grounding. Trigger on phrases like "what do you know about X", "let's talk about X", "explain X", "I'm trying to understand X", "discuss X with me", "what does the KB have on X", "help me think through X", "tell me about X", or any conversational question about a concept or idea. Also trigger when the user asks a question that could be answered from stored knowledge, even if they don't explicitly say "knowledge base". This is the default skill for knowledge-grounded conversation — prefer it over answering from memory alone whenever the KB might have relevant content. Do NOT trigger for: deep research reports (use keine-research), adding new entries (use keine-update-entries), or creating maps (use keine-update-maps).
A conversational skill for exploring ideas and answering questions using the local knowledge base. The KB is the primary source of truth — answer from it, be honest about its limits, and offer to fill gaps with web research when the user wants more.
This is a conversation, not a report. Stay concise and direct. The goal is to give the user a grounded, honest answer — citing what the KB actually says, noting what it doesn't cover, and offering to go further if they want.
Before searching, make sure you understand what the user actually wants:
If the question is ambiguous, make a reasonable interpretation and proceed — you can clarify inline as you answer.
Search efficiently. Start broad, narrow as needed:
# Check for relevant topic maps (fastest orientation)
ls docs/maps/
# Check tags for the topic area
ls docs/tags/
# Full-text search for keywords
grep -ri "keyword" docs/ -l
# Read a tag index to find related entries
cat docs/tags/<slug>.md
Read the most relevant files. For a focused question, 2–5 entries is usually enough. For a broad topic, read the map file first — it gives the quickest overview and links to key entries.
As you read, track:
Compose your answer from what you found. Keep it conversational:
Don't pad. A sharp 3-paragraph answer is better than a bloated 10-paragraph one. If the KB has a lot of depth, offer to go deeper on a specific facet rather than dumping everything at once.
After answering, briefly note what's missing if it's relevant to the user's question:
"The KB doesn't have much on [aspect]. Want me to search the web and add entries for it?"
Only offer this when the gap is meaningful. Don't prompt for web research if the KB already answered the question well.
If the user says yes, move to Step 5.
Search the web to fill the gaps. For each significant finding:
keine-update-entries to create a new KB entry if the knowledge is worth keepingkeine-update-mapsscripts/maintain_tags.pyCommit new entries:
git add docs/ docs/tags/
git commit -m "docs: add <topic>"
keine-research (heavyweight, multi-phase, produces a file in reports/)keine-update-entrieskeine-update-maps