SuperAPI AI Directory

wiki-query

General & other

View repository

Updated 10-15 23:45

About this skill

Wiki Query

Ask a question. Read the wiki. Synthesize with citations. Offer to file the answer back.

Pre-condition

Find SCHEMA.md (search from cwd upward, or ~/wikis/). If not found, tell the user to run wiki-init first. Read it to get wiki root path and cross-reference convention.

Link style. Read link_style: from SCHEMA.md. If the field is missing or <wiki-root>/config/link-style.md does not exist, default to obsidian ([[slug]]). Otherwise, read <wiki-root>/config/link-style.md for the emit and parse rules. Follow cross-references in pages using the ## Parse rule. When you write the answer back as a wiki page (step 4), emit citations and cross-references using the ## Emit rule.

Process

1. Read wiki/index.md first

Regenerate the index so it reflects the current pages — python bin/generate-index.py — then scan the full index to identify which pages are likely relevant. Do NOT answer from general knowledge — the wiki is the source of truth here, even if you think you know the answer.

2. Read relevant pages

Read the identified pages in full. Follow one level of cross-reference links (matched via the ## Parse rule from config/link-style.md) if they point to pages that seem relevant to the question.

3. Synthesize the answer

Write a response that:

  • Is grounded in the wiki pages you read
  • Cites inline using a cross-reference (in the wiki's link_style) for every claim sourced from a specific page
  • Notes agreements and disagreements between pages
  • Flags gaps: "The wiki has no page on X" or " doesn't cover Y yet"
  • Suggests follow-up sources to ingest or questions to investigate

Format for the question type:

  • Factual → prose with citations
  • Comparison → table
  • How-it-works → numbered steps
  • What-do-we-know-about-X → structured summary with open questions

4. Always offer to save

After answering, say:

"Worth saving as wiki/pages/<suggested-slug>.md?"

If yes:

  • Write the page with full frontmatter: title, category: Analyses (or the wiki's equivalent), summary (one-line), tags: [query, analysis], sources: [all cited slugs], created: <today>, updated: <today>
  • Regenerate the index — python bin/generate-index.py — do not hand-edit index.md
  • Record the operation (per SCHEMA's Operation Log & Commit Convention):
    • Git wiki: suggest a commit (default docs: for a saved query, plus the trailer); commit on confirmation.
      docs: add analysis — <question summary>
      
      Wiki-Op: query
      
    • Non-git wiki: append to wiki/log.md:
      ## [<today>] query | <question summary>
      Filed as: <slug reference to the new slug, in the wiki's link_style>
      

If no:

  • Git wiki: nothing changed, so there is nothing to commit — the query is not logged.
  • Non-git wiki: append to wiki/log.md:
    ## [<today>] query | <question summary>
    Not filed.
    

Common Mistakes

  • Answering from memory — Always read the wiki pages. The wiki may contradict what you think you know, and that contradiction is valuable signal.
  • Skipping the save offer — Good query answers compound the wiki's value. Always offer.
  • No citations — Every factual claim should trace back to a slug reference (in the wiki's link_style).