Borrowing it
Nothing to install: this file belongs to patrickjaja/claude-desktop-extra. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/patrickjaja/claude-desktop-extra/master/.claude/skills/architecture/SKILL.mdgit clone --depth 1 https://github.com/patrickjaja/claude-desktop-extraWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/patrickjaja/claude-desktop-extra/architecture)<a href="https://agentmods.dev/skills/patrickjaja/claude-desktop-extra/architecture"><img src="https://agentmods.dev/badge/skills/patrickjaja/claude-desktop-extra/architecture.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00095 | $0.01483 |
| Opus 5 | $0.00048 | $0.00741 |
| Sonnet 5 | $0.00019 | $0.00297 |
| Haiku 4.5 | $0.00010 | $0.00148 |
Grade A, and why
architecture scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Architecture - claude-desktop-extra
Repackages Anthropic's official Claude Desktop Linux .deb for the distros Anthropic does not ship (Arch via our own pacman repo, Fedora/RHEL, NixOS, AppImage) and for our own Debian/Ubuntu .deb, while layering a small set of Linux-only value-adds that the official build does not provide. See /linux for compat specifics.
What we start from: the official Linux .deb
Anthropic publishes an official Claude Desktop build for Linux as a .deb in an apt repo (https://downloads.claude.ai/claude-desktop/apt). It bundles its own Electron (42.5.1) and ships a native Cowork VM backend (cowork-linux-helper + virtiofsd + smol-bin + QEMU/OVMF; requires /dev/kvm). The official build natively supports Chat, Code, Cowork, Computer Use is in beta upstream, and reads managed config from /etc/claude-desktop/managed-settings.json. We do not patch the Windows MSIX any more - that pipeline is gone.
What this repo does: ingest -> patch -> repackage
Ingest: download the official .deb (apt repo, or --deb PATH/--version X) -> verify GPG + SHA256 -> dpkg-deb -x -> locate usr/lib/claude-desktop/resources/app.asar -> asar extract.
Patch: run scripts/apply_patches.py (the orchestrator) over the extracted bundle. 48 surgical JS patches (patches/{linux,community,core}/*.nim, compiled native binaries, regex on minified JS) apply Linux-only fixes and our value-adds. The official build re-minifies between releases, so patterns use [\w$]+ wildcards anchored on stable strings (feature names, log messages, process.platform==="darwin"), count EXPECTED_PATCHES, and quit(1) on any miss. .upstream-version records the last validated version.
Repackage: asar pack (preserving app.asar.unpacked, which carries the official build's pre-built native modules - we no longer rebuild node-pty) into a tarball, then build the pacman package/our-own-deb/rpm/AppImage/Nix from it.
Several patches that used to enable Cowork on Linux (the "cowork-wiring" cluster) were removed once the official build started shipping Cowork natively; what remains of that area is a small number of regression guards that assert the upstreamed native-Linux behavior is still present and fail loud if Anthropic ever drops it.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 53 lines · 95 tokens per session scan A 6f307c01b872
architecture is a skill published in the GitHub repository patrickjaja/claude-desktop-extra (620 stars, last pushed 2d ago), licensed MIT. It adds 95 tokens to every session and 1,483 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
bastra-recall
Persistent external brain for ChatGPT, Codex, Claude, and other MCP clients — documents (PDFs, contracts, scans with OCR), personal facts (appointments, decisions, items, amounts), AND code lessons / preferences / project topology. USE PROACTIVELY in three modes. (1) RECALL — whenever the user asks about anything from…
bastra-recall
Proactive private local memory for ChatGPT and Codex — recall before acting, save durable rules, lessons and decisions without being asked. Requires the local bastra-recall MCP server installed by bastra install codex.
calibre-distill-topic
Synthesize ONE topic across MULTIPLE Calibre books into a single, concept-keyed agent skill — original multi-source authorship (decision frameworks, cross-source config tables, an explicit where-authors-agree/disagree section), with a full ISBN bibliography that doubles as a live-source binding for later deep dives.…
calibre-mcp-setup
Install and verify the calibre-mcp server, end to end. Guides an agent from 'Calibre is installed' to a verified working MCP connection: preflight (Node, calibredb, Content Server — macOS/Windows/Linux commands), client-specific install (Claude Code plugin or claude mcp add, Claude Desktop .mcpb or JSON config, Cowork…
calibre-study-notes
A tool that creates study notes from books in a Calibre library for a human learner. It uses questions, recall exercises, source pointers, and optional Anki export rather than only providing a summary.
himalayas-employer
AI hiring assistant for Himalayas. Post remote jobs, manage company profiles, source talent, message candidates, and benchmark salaries using the Himalayas MCP server.