Wiki Log#

[2026-04-07] create | Wiki initialized#

Wiki structure created: CLAUDE.md schema, wiki/index.md, wiki/log.md, directory scaffolding. Ready for first ingest.

[2026-04-07] ingest | Scion Documentation#

Source: https://googlecloudplatform.github.io/scion — multi-agent orchestration testbed by Google Cloud Platform. Created 13 wiki pages:

[2026-04-07] ingest | Kiro Autonomous Agent#

Source: https://kiro.dev/autonomous-agent/ — AWS’s frontier agent for autonomous development tasks. Created 6 wiki pages, updated 1:

[2026-04-07] ingest | Claude Code Documentation#

Source: https://code.claude.com/docs/en — Anthropic’s agentic coding tool. Created 4 wiki pages, updated 2:

[2026-04-07] update | Claude Code Documentation (deep ingest)#

Discovered raw .md endpoints on the docs site. Re-ingested overview, how-claude-code-works, memory, sub-agents, permission-modes, hooks, and skills pages with full content. Updated:

  • claude-code-docs — comprehensive takeaways from actual doc content
  • claude-code — full architecture (agentic loop, tools, memory, permissions, extensibility)
  • mcp-protocol — richer detail on Claude Code’s MCP usage

[2026-04-07] ingest | Anthropic Skills Repository & Agent Skills Spec#

Source: https://github.com/anthropics/skills + https://agentskills.io/specification — open standard for agent skills and Anthropic’s reference implementation. Created 2 wiki pages, updated 3:

[2026-04-08] ingest | The Ten Pillars of Agentic Skill Design#

Source: ~/Documents/2-Areas/agentic-skills-research/agentic-skills-best-practices.pdf (.md version) — research paper by Ian Forster proposing a ten-pillar framework for skill design. Created 3 wiki pages, updated 2:

[2026-04-08] ingest | LLM Wiki (Karpathy)#

Source: https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f — the foundational idea file for the LLM Wiki pattern. This wiki is a running instance of this pattern. Created 3 wiki pages:

[2026-04-08] ingest | Fabric#

Source: https://github.com/danielmiessler/fabric — open-source framework with 251+ curated AI prompt patterns and composable strategies. Created 3 wiki pages, updated 2:

[2026-04-08] ingest | Personal AI Infrastructure (PAI)#

Source: https://github.com/danielmiessler/Personal_AI_Infrastructure — personalized AI platform built on Claude Code with memory, skills, goals, and continuous learning. Created 2 wiki pages, updated 1:

[2026-04-08] query | Key Insights: The Agentic AI Landscape#

Synthesized 10 key insights from all 8 sources in the wiki. Filed as analysis: key-insights-agentic-landscape. Covers: three architectural philosophies, the autonomy-interaction spectrum, emerging open standards, progressive disclosure consensus, memory as unsolved frontier, git as universal coordination, skill hierarchy crystallization, security trust models, the personal AI vision, and what’s missing.

[2026-04-08] ingest | How to Evaluate AI Agent Skills Without Relying on Vibes#

Source: Article by JP Caparas — practical guide to skill evaluation with three-tier framework and economics. Created 2 wiki pages, updated 1:

[2026-04-09] ingest | strAIght talk: AI Tips for Amazonians#

Source: Podcast notes from raw/human/ai-technique-podcast.md — Amazon employees sharing AI workflow tactics. Created 1 wiki page:

[2026-04-09] update | Overview#

Updated overview.md with current stats (10 sources, 18 concepts, 10 entities), added “Key Themes” section synthesizing cross-cutting patterns, added ai-technique-podcast to practitioner insights.

[2026-04-09] ingest | Skills Pipeline (Sleestk)#

Source: https://github.com/Sleestk/Skills-Pipeline — chained skill pipelines for YouTube production, SaaS development, and Obsidian. Created 1 wiki page:

  • Source: skills-pipeline-sleestk
  • Key insight: Skills as pipelines — each stage’s output is the next stage’s input. Concrete implementation of progressive disclosure and context management recipes. Ships with inline test prompts.

[2026-04-09] query | Ten Pillars Evidence Map#

Synthesized evidence from all 11 wiki sources supporting each pillar of the Ten Pillars framework. Filed as analysis: ten-pillars-evidence-map. Strongest pillars: Scope (SRP), Modularity, Context Management. Weakest: Versioning, Testing. Identified 4 gaps for v3.

[2026-04-09] query | Cross-Source Theme Analysis#

Identified 8 common themes across all 11 sources with evidence tables and strength rankings. Filed as analysis: cross-source-themes. Strongest: “Context is king” (9/11), “Composition over monoliths” (8/11). Weakest: “Evaluation is weakest link” (4/11 — but universally acknowledged as a gap).

[2026-04-09] query | How to Eval a Skill#

Practical guide mapping Anthropic’s prompt eval methodology onto skills. Filed as analysis: how-to-eval-a-skill. Covers: 5 testable surfaces (routing, tool selection, process, side effects, output quality), 3 tiers, pass@k, eval.yaml format, Claude Code hook integration, CI/CD pipeline.

[2026-04-09] ingest | Anthropic Eval Guide + Promptfoo#

Two eval resources ingested. Created 3 wiki pages:

  • Source: anthropic-eval-guide — canonical eval methodology (SMART criteria, eval types, “volume over quality”)
  • Source: promptfoo — open-source eval CLI (YAML test cases, CI/CD, red teaming)
  • Entity: promptfoo — the closest existing tool to a turnkey skill eval pipeline

[2026-04-09] ingest | Paperclip#

Source: https://github.com/paperclipai/paperclip — open-source orchestration for zero-human companies. Created 2 wiki pages, updated 1:

[2026-04-10] ingest | Spec Kit#

Source: https://github.com/github/spec-kit — GitHub’s spec-driven development toolkit. Created 2 wiki pages:

  • Source: spec-kit
  • Entity: spec-kit
  • Key insight: SDD is progressive disclosure applied to development — each step produces a focused artifact that feeds the next. Agent-agnostic (30+ agents), with 50+ community extensions. Operates at the methodology layer between organizational orchestration (Paperclip) and tool-level execution (Claude Code).

[2026-04-10] ingest | BMad Method#

Source: https://github.com/bmad-code-org/BMAD-METHOD — AI-driven agile framework with scale-adaptive intelligence. Created 2 wiki pages:

  • Source: bmad-method
  • Entity: bmad-method
  • Key insight: “Expert collaboration over autopilot” — agents guide your thinking rather than replacing it. Scale-adaptive planning adjusts depth to project complexity. Party Mode enables multi-persona dialogue in a single session.

[2026-04-11] ingest | NotebookLM Notes Guide#

Source: https://medium.com/@stevenbjohnson/getting-the-most-out-of-notes-in-notebooklm-d9d70316b780 — how-to guide by Steven Johnson on NotebookLM’s notes system. Created 3 wiki pages, updated 2:

  • Source: notebooklm-notes-guide
  • Entities: notebooklm, steven-johnson
  • Updated: context-management — added provenance-as-context-metadata section, NotebookLM’s 5K word limit as concrete context constraint
  • Updated: llm-wiki-pattern — added contrast section comparing NotebookLM’s session-oriented approach to the cumulative wiki pattern
  • Key insight: NotebookLM validates the “human curates, AI processes” principle from a different angle — interactive exploration with provenance tracking vs. structured ingest with automatic cross-references.
Backward Ten Pillars Agentic Skill Design