ai-topics

Comprehensive wiki health checking, maintenance, and remediation — duplicate detection, entity dedup/disambiguation, link auditing/fixing, page splitting, source-linking lint, wikilink remediation, tag taxonomy audit and normalization, pre-commit enforcement, language enforcement (JP→EN bulk translation, detection regex, cron-assisted migration), and decision-matrix-driven cleanup.

Comprehensive wiki maintenance: daily structural health checks (index reconciliation, log separator fixes, pipeline watchdog alerts), comparison page updates (adding items to multi-section comparison tables), and page relocation (moving/renaming pages while maintaining link integrity).

Pre-flight checklist and procedures for archiving, deleting, or migrating Hermes skills. Prevents accidental removal of cron-referenced skills. Covers the 3-layer skill structure and config.yaml management. Includes skill inventory management, promotion workflows, and archival conventions.

Karpathy's LLM Wiki: build/query interlinked markdown KB.

Deep analysis of blog authors' recent thoughts, philosophy, and positions. Goes beyond entity page creation to extract cited ideological positions, track thought evolution, and enable ongoing RSS monitoring for thought updates.

Query the blogwatcher-cli SQLite database for RSS scan results. Use pre-verified column names and query templates to avoid errors.

Synthesize perspectives from multiple opinion leaders on a shared topic into a cross-referenced wiki index page. Use when the user wants to consolidate different viewpoints on the same concept.

Daily RSS scan triage workflow — scan blogs, triage articles, ingest into wiki, commit changes

Knowledge consolidation cycle that analyzes collected articles and updates the wiki with significant findings while enforcing duplicate checks.

Enrich person entity pages using Grokipedia (AI-generated encyclopedia with 6M+ articles). Search for people by name, extract structured biographical data, and merge into wiki entity pages.

Systematic depth analysis framework for opinion leaders tracking research completion across 4 layers - L1 Profile, L2 Timeline, L3 Thought Analysis, L4 Ongoing Monitoring

Group raw articles or newsletter-ingest checkpoints by semantic topics, assess wiki value against existing coverage, and recommend actions

Create or enrich wiki entity and concept pages from raw articles, multi-source web research, newsletters, and tool/project investigations. Supports both article-driven enrichment and from-scratch entity page creation via website + docs + X/Twitter + GitHub/NPM research.

Upgrade bio-only blogger entity pages to comprehensive thought analysis format

Umbrella for all wiki ingestion pipelines — newsletter ingest, blog ingest, active knowledge crawl, arXiv paper pipeline, OpenAI blog ingestion, dreaming knowledge consolidation, and pipeline troubleshooting.

Enrich skeleton X/Twitter account entity pages to full quality (8-15KB), matching antirez-com.md / simon-willison.md depth.

Scheduled autonomous wiki crawling to discover and document trending AI/ML developments

Research and compare AI agent harnesses (coding/automation agents like Claude Code, Pi, OpenCode, OpenClaw, Hermes Agent) with model compatibility analysis. Covers the "Harness Effect" — how the same model performs 5-40 percentage points differently depending on the harness.

Search arXiv papers by keyword, author, category, or ID.

Write and iteratively refine Japanese-language technical blog posts — wiki research first, humanizer review, multi-round user feedback integration.

Write original blog posts for the AI topics blog. Save to blog/ directory (NOT wiki/raw/articles/).

Monitor blogs and RSS/Atom feeds for updates using the blogwatcher-cli tool. Add blogs, scan for new articles, track read status, and filter by category.

ユーザーから「詳細すぎる／特殊的すぎる／視野を広げて」というフィードバックを受けた際に発火。Gwernの「seeing through / unseeing」フレームワークを用いて、還元主義から抽象化への認知シフトを行い、より広い視点で再分析する。

Deliver Hermes AI Topics reports via Cloudflare Email Service. Setup, cron config, and troubleshooting for email channel.

Hermes cron job lifecycle management — creation, model configuration, manual testing, and execution monitoring.

Policy for ingesting official documentation pages (product docs, developer guides, API references) into the wiki. Prioritizes creating a comprehensive overview guide of the subject over mechanically mirroring the source structure.

Effective delegate_task goal patterns for 7 categories of autonomous work — QA overnight, research, migrations, batch processing, docs, perf, and specs. Based on @0xsero's goal delegation best practices.

Complete guide to using and extending Hermes Agent — CLI usage, setup, configuration, spawning additional agents, gateway platforms, skills, voice, tools, profiles, and a concise contributor reference. Load this skill when helping users configure Hermes, troubleshoot issues, spawn agent instances, or make code contributions.

Gwern-inspired quality techniques for Hermes AI reports — Anti-Examples stripping, Manual of Style (MoS), Atomic Snippets multi-level abstraction, Generate-Rank-Select iteration, and Engram knowledge pathways. Improves report tone, consistency, depth, and usefulness.

Humanize text: strip AI-isms and add real voice.

Pitfalls, examples, and edge cases for Hermes Kanban workers. The lifecycle itself is auto-injected into every worker's system prompt as KANBAN_GUIDANCE (from agent/prompt_builder.py); this skill is what you load when you want deeper detail on specific scenarios.

Location intelligence — geocode a place, reverse-geocode coordinates, find nearby places (46 POI categories), driving/walking/cycling distance + time, turn-by-turn directions, timezone lookup, bounding box + area for a named place, and POI search within a rectangle. Uses OpenStreetMap + Overpass + OSRM. Free, no API key.

Extract text from PDFs/scans (pymupdf, marker-pdf).

Naming convention for raw article files in wiki/raw/articles/ — always use the actual publication date, not the date of ingestion.

Execute plans via delegate_task subagents (2-stage review).

Regularly scan RSS feeds, blogwatcher DB, and raw articles for trending AI/ML topics. Synthesize findings into a Japanese-language report with top 5-8 topics, source links, and wiki-action recommendations.

When creating or ingesting wiki pages that compare 3+ items (tools, models, frameworks, platforms, etc.), route them to wiki/comparisons/ — NOT wiki/concepts/. This prevents multi-item comparison pages from accumulating in the concepts/ directory where they're hard to find and require cleanup.

Create a new wiki concept page from scratch via web research (not from raw articles), OR rebuild/expand an existing concept page with fresh research. Includes explicit differentiation from related pages.

Generate a Japanese-language daily wiki change report summarizing today's modifications. Handles the case where wiki_report.py doesn't exist by using find-based fallback.

Daily wiki structural maintenance patterns — index reconciliation, log separator fixes, pipeline watchdog alerts, and auto-fix verification for the ai-topics wiki.