orriborri/pasta/llm-wiki

Claude Code and agent skill by orriborri/pasta

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Maintain this vault's Pasta-backed LLM wiki and curate JDex notes

Install

npx vskill@latest install orriborri/pasta/llm-wiki

SKILL.md

LLM Wiki Maintenance

The compiler is vault.py in the Pasta repo checkout (~/code/pasta/llm-wiki/vault.py). The vault root comes from ~/.pasta/config.toml → [general].vault_path; generated proposals live under .llm-wiki/generated/NN.NN/ using the private JDex registry.

Pick the right tool

  • Searching or reading notes: Pasta search_knowledge, or read the files directly. Never vault.py.
  • Creating or editing an ordinary note: write the file (or use the Obsidian CLI). Never vault.py.
  • Saving a new personal statement: the raw-inbox-memory skill writes it to 0. Inbox/Raw/. Never write it into the wiki.
  • Wiki upkeep: vault.py, following the cycle below.

Wiki upkeep cycle

Follow ~/code/pasta/llm-wiki/workflows/incremental.md. The rules that matter:

  1. vault.py cursor, then fetch changes with Pasta get_changes.
  2. vault.py plan to get jobs; use their enabled JDex routing entries and prompts/maintain.md. Each patch and frontmatter declares one matching jd_id. Skip uncertain routing with a review reason.
  3. vault.py validate every patch. If validation fails, stop: do not apply, do not advance, and report the validator's message verbatim.
  4. vault.py apply only validated patches.
  5. vault.py advance --changes <page.json> only after every job on that page is applied. When stopping with has_more still true, advance completed pages only. Each job needs a verified apply or explicit skip receipt.

The nightly pasta-llm-wiki.timer runs this cycle unattended via ~/code/pasta/llm-wiki/run-upkeep.sh, with logs under ~/.local/state/pasta/llm-wiki/runs/.

Every factual claim in wiki prose cites its evidence as pasta:evidence:<record_id>. Generated prose is context, not evidence.

JDex curation

For an explicitly requested canonical-note update, retrieve its approved JDex entry and the relevant generated proposal. Verify claims through Pasta evidence, then propose a cited update to that ID's existing canonical note. Preserve human content and source files. Uncertain destinations remain review items. Placement uses the existing area/category/ID hierarchy, not model-created categories.

Initialize the toolkit with init --wiki ROOT --registry PRIVATE_REGISTRY. Publication requires an approved or management entry with publication_enabled set to true. Registry edits invalidate outstanding plans; replan before writing.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/orriborri/pasta/llm-wiki/badge)](https://verified-skill.com/skills/orriborri/pasta/llm-wiki)

Something wrong with this skill or its page? Tell us.