agentleFS
Edzard9Edzard9/stash/skills/vault-router/SKILL.md

vault-router

Skill0 starsChanged 8 days agoMITView on GitHub

Entry point for the skill-vault collection. Use FIRST whenever a user makes a task request and any skill-vault skills are installed: read the request, analyze what the actual task and deliverable are, then route to the one matching vault skill and follow it. Also use when the user asks which skill can do something, or when a request does not obviously belong to any single skill.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Interactive figure vs video: If the user wants to click and explore, it is interactive-figure.

    line 68
  2. Distinctive:

    The right skill is not installed: Do not invent it and do not fake its workflow.

    line 72
  3. Distinctive:

    Do not route based on one keyword while ignoring the deliverable.

    line 85
  4. Distinctive:

    Do not load two full skills "just in case".

    line 86
  5. Distinctive:

    Do not mention this router's mechanics at length.

    line 88
  6. Distinctive:

    When a skill is added to or removed from the vault, add or remove its row in the catalog above in the same change.

    line 93

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Vault Router

You are the front door of the skill-vault. The user should never have to know skill names. They describe what they want; you figure out which vault skill does it, load that skill, and follow it.

How to route

  1. Read the whole request first. Identify the task by its verb, its object, and above all its final deliverable (a file, a video, a PR, a deployed service). The deliverable is the strongest routing signal, stronger than any single keyword.
  2. Match against the catalog below. Each row lists a skill, what it produces, and typical signal phrases in English and Indonesian.
  3. Check it is actually installed. Route only to skills present in the current environment (your available-skills list, or sibling folders in the skills directory). The catalog lists the whole vault; the user may have installed only some of it.
  4. Load and follow. Open the matched skill's SKILL.md in full and do the task by its instructions. Tell the user in one short line which skill you picked, for example: "Routing to lab-report for this."
  5. Stay in the routed skill until the task is done or it clearly hands off to another one.

Catalog

SkillProducesSignals (EN / ID)
lean-codeCode written or refactored to be concisecode, refactor, clean up / rapikan kode, bikin ringkas
oss-contributorA contribution to someone else's repo: screened claim, PR, disclosurecontribute, external PR, good first issue / kontribusi, klaim issue, PR ke repo orang
small-briefsA big build split into small verified briefs for a coding agentbreak this down, briefs / pecah tugas, delegasikan, brief untuk opencode
cloudflare-workerCloudflare Workers + D1 apps and fixescloudflare, wrangler, D1, worker / deploy worker, migrasi D1
laravel-patternsLaravel architecture patterns, adapted from ECClaravel, eloquent, artisan / pola laravel, bikin fitur laravel
laravel-securityLaravel security checklist and hardening, adapted from ECClaravel security, sanctum, csrf / keamanan laravel, audit keamanan
react-patternsReact 18/19 patterns and hooks discipline, adapted from ECCreact, hooks, useEffect / komponen react, state react
nextjs-turbopackNext.js 16+ and Turbopack guidance, adapted from ECCnext.js, turbopack, proxy.ts / next dev lambat
ui-craftUI and interface design workdesign, UI, layout, landing page / desain tampilan, bikin UI
title-cardTitle cards and thumbnails as PNG imagesthumbnail, title card, cover image / thumbnail, cover video
lab-reportA practicum report in the owner's house formatlab report, practicum / laporan praktikum, hasil pengamatan
interactive-figureA clickable HTML figure for a study conceptinteractive diagram, simulation / figur interaktif, diagram bisa diklik, simulasi konsep
anatomyCrafted isometric SVG figures of a concept as a physical object, adapted from wheresryan22/anatomyisometric figure, svg illustration / figur isometrik, gambar penjelasan
cisco-ios-patternsCisco IOS review patterns and change windows, adapted from ECCcisco, ios, access-list / konfigurasi cisco, wildcard mask
net-labNetworking lab configs and validation, module-faithfulVLAN, routing, Packet Tracer, subnet / jarkom, konfigurasi switch, praktikum jaringan
study-notesStudy notes and flashcards from slides or PDFssummarize, flashcards, notes / rangkum materi, catatan belajar, flashcard
html-videoA rendered MP4 video from a text briefvideo, MP4, shorts, reels, explainer / buatkan video, video penjelasan
manim-explainerMath and science animations in Manimmanim, math animation / animasi matematika, animasi rumus
vps-deployAn app deployed to a VPS: systemd, Tailscale-only, previewdeploy, VPS, server, systemd / deploy ke VPS, online-kan, pasang di server
repo-auditA security audit of the user's own public reposaudit my repos, leaked secrets / audit repo, cek kebocoran, secret bocor
homelab-vlan-segmentationHome network VLAN segmentation, adapted from ECChome vlan, iot isolation / vlan rumah, pisahkan iot, guest wifi
caption-craftSocial media captions and carousel text, ID/ENcaption, carousel, post text / caption IG, teks carousel
proofread-idFormal Indonesian proofreadingproofread, spelling / cek ejaan, rapikan bahasa, periksa tulisan
councilA hard decision pressure-tested by five advisor lenses with anonymous cross-review and one synthesized verdictdecide, should I X or Y, which option, pressure-test / pilih yang mana, minta pendapat kedua, uji keputusan, bantu putuskan

Edge cases

  • Two skills fit. Route to the one that owns the FINAL deliverable as primary; use the other as a supporting reference and say so in one line. Example: "a VLAN explainer video" routes to html-video (deliverable: MP4) with net-lab consulted for technical correctness. "Deploy my Workers app" routes to cloudflare-worker for the app and vps-deploy only if a server is actually involved.
  • Interactive figure vs video. If the user wants to click and explore, it is interactive-figure. If they want a file that plays, it is html-video or manim-explainer. When genuinely ambiguous, this is the one case worth a single short question.
  • The right skill is not installed. Do not invent it and do not fake its workflow. Name the skill, point to its path in the skill-vault repo, and either use the closest installed skill or do the task unaided, saying which you are doing.
  • Nothing matches. Say plainly that the vault has no skill for this yet. Do the task normally if you can. If the request smells repeatable, close by suggesting it as a new vault skill: a name (English, cool, simple), its shelf, and one line on what it would do. The vault grows exactly this way.
  • The user names a skill directly. Skip routing and follow it.

What not to do

  • Do not route based on one keyword while ignoring the deliverable.
  • Do not load two full skills "just in case"; route, then chain only on a real hand-off.
  • Do not mention this router's mechanics at length. One routing line to the user is enough; then get to work.

Maintaining this file

When a skill is added to or removed from the vault, add or remove its row in the catalog above in the same change. A router pointing at skills that do not exist is worse than no router.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.