本文へ移動
cccskills
無料GitHub で公開

Wiki Ingest

Ingest raw sources into the structured wiki — "ingest", "add this to my wiki", new files in raw/, clippings, sources, ideas, nightly ingest. Produces pages with YAML frontmatter and a mandatory Counter-Arguments section.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md6.1 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Wiki Ingest

Purpose

Convert raw, unstructured sources into structured wiki pages stored in MemPalace. Each page has YAML frontmatter, mandatory Counter-Arguments section, bidirectional [[wikilinks]], and explored: false until human verification.

This is one of 4 wiki skills (ingest, query, lint, explore) that together implement the Brandvoice pattern on top of MemPalace.

When To Invoke

  • New files appear in raw/ that don't have a corresponding wiki entry
  • User says "ingest", "add this to my wiki", or drops a file into raw/
  • Nightly scheduled cycle (configured via /mxm-tasks)
  • After any web research session (research agents auto-trigger)

Inputs

  • Source files in raw/:
    • raw/clippings/ — web clips, article dumps
    • raw/ideas/ — freeform notes, brain dumps
    • raw/sources/ — PDFs, papers, transcripts
    • raw/x-archive/ — social exports

Workflow

  1. Scan raw/ for files not yet referenced by any MemPalace drawer (compare against mempalace_search results filtered by source path).
  2. For each unprocessed source:
    • Detect type: tweet, article, PDF, transcript, note, bookmark, paper
    • If URL: fetch full content via WebFetch
    • Extract: key claims, concepts, entities, dates, counter-arguments, citations
    • Apply 80% relevance filter — skip if signal is weak (vague, redundant, off-topic)
  3. Create source summary page in MemPalace:
    • Wing: wing_knowledge
    • Room: sources
    • Drawer: page body with YAML frontmatter (see schema below)
  4. Create or UPDATE concept/entity pages (never overwrite — merge):
    • Concepts → wing wing_knowledge, room concepts
    • Entities (people, tools, companies) → wing wing_knowledge, room entities
  5. Add bidirectional [[wikilinks]] between all related pages (use mempalace_kg_add to record relationships in the graph layer).
  6. Set explored: false on all new/updated pages.
  7. Update wiki index: append TLDR entries to MemPalace drawer wing_knowledge/index/index-md.
  8. Append to wiki log: timestamp, source name, pages created/updated → MemPalace drawer wing_knowledge/log/log-md.

Page Schema (MemPalace drawer frontmatter)

---
title: ""
created: YYYY-MM-DD
updated: YYYY-MM-DD
tags: []
sources: []                  # paths to raw/ files
confidence: high | medium | low
explored: false              # AI-generated → human verification needed
startup: null                # set if scoped to a specific startup
---

Mandatory Page Sections (in order)

  1. TLDR — 2–3 sentence summary
  2. Main Content — H2/H3 structured body
  3. Counter-Arguments — data gaps, opposing evidence, caveats (NEVER skip)
  4. Related — [[wikilinks]] to other pages (≥2 inbound for new pages)
  5. Sources — links back to raw/ files

Quality Rules

  • Apply 80% relevance filter — skip sources with weak signal
  • Every page MUST have Counter-Arguments section
  • Never duplicate pages — merge into existing pages when topic overlaps
  • Confidence defaults to medium unless source is official/verified
  • Naming: lowercase kebab-case (agent-memory-patterns)
  • One concept per page — split any page over 500 lines

MemPalace Integration

mempalace_add_drawer({
  wing: "wing_knowledge",
  room: "sources" | "concepts" | "entities" | "syntheses",
  content: "<full-page-with-frontmatter>",
  source_file: "raw/sources/<original>.pdf",
  added_by: "wiki-ingest"
})

mempalace_kg_add({
  subject: "<entity-or-concept>",
  predicate: "mentioned_in" | "related_to" | "contradicts" | "extends",
  object: "<related-entity-or-concept>",
  source_closet: "<drawer-id>"
})

Output Format

Wiki Ingest Report:
Sources processed: [N]
Skipped (low signal): [N]
Pages created:
  - sources/[slug] (confidence: medium, explored: false)
  - concepts/[slug] (confidence: medium, explored: false)
  - entities/[slug] (confidence: high, explored: false)
KG triples added: [N]
Index entries appended: [N]
Log appended: 1 entry @ [ISO timestamp]
Next: review pending pages with `/mxm-wiki explore`
Confidence: 🟢 HIGH (relevance ≥80%, all pages have Counter-Arguments)

Handoff

  • Pages created → notify wiki-explore for the next verification pass
  • Contradictions detected during ingest → escalate to wiki-lint
  • Pages tagged with regulated topic (GDPR/HIPAA/etc.) → notify compliance-officer for review
  • New concepts that should propagate to brand voice → notify brand-guardian

Maxim Behavioral Layer

  • Confidence tagging: 🟢 HIGH when relevance ≥80% and all pages structured correctly | 🟡 MEDIUM when partial structure | 🔴 LOW when relevance filter rejected most sources
  • Frameworks: LLM Wikid pattern (sources → structured wiki with bidirectional links + explored: false default + Counter-Arguments)
  • Ethics gate: never ingest sources containing PII without data-privacy-officer review

Source References

  • FINAL_RELEASE_CHANGES/Brandvoice-wiki-example.md — original pattern
  • documents/reference/AGENTS.md — agent usage guide for the wiki layer
  • .claude/skills/memory-palace/SKILL.md — MemPalace integration patterns

Copyright (c) 2026 iSystematic Inc. Maxim product. BSL 1.1.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

AI video and image prompt engineering — video generation, image generation, product video, social media video, cinematic direction, animation, Seedance, Higgsfield, GPT-Image-2. Applies 15 Maxim-authored cinematic styles and the atomic prompt schema.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Apply a named behavioral science framework to any task — persuasion, nudge design, habit formation, conversion, influence, negotiation, cognitive bias, choice architecture, decision framing. The framework backbone every other Maxim skill loops into for psychological validation.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Founder and CEO automation — morning standup, overnight cycle, burn rate, runway, pipeline review, investor and fundraising prep, executive summary, daily digest. Operates on the .mxm-executive-summary context.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Tear down a company on request — "do a teardown of", "break down this company", "analyze this competitor's business". Fixed-template output: business model, how they make money, positioning, and where they are exposed.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Track what a competitor actually shipped — "watch this competitor", "track what they ship", "what changed at". Compares against the last snapshot and reports only real moves: launches, pricing changes, positioning shifts.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Regulatory review across 14 frameworks — compliance review, privacy impact assessment, regulatory gap analysis, data subject rights, consent mechanisms, audit trail, AI ethics review, data flow mapping. Covers GDPR, PIPEDA, HIPAA, PCI-DSS, SOC2, UAE-PDPL, EU AI Act and more.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

DrNabeelKhan のスキルをすべて見る

このスキルの問題を報告する