amaniagent/skills
amaniagent · open agent tooling

skills

A skill is executable trust — you install someone's instructions and their scripts into an agent that can touch your files, your keys, your machine.

A supply chain needs a gate. This is a curated catalog of skills reviewed for what they contain, how they behave, and what they ask an agent to access.

03 / ACCESSmachine
boundary
02 / REVIEWhuman
audit
01 / SOURCEread-only
by design
no target scripts executed

Categories

10 categories · 46 skills
7 SKILLS

security

answer-verifier · dependency-auditor · mcp-auditor · prompt-injection-detector · repo-auditor · settings-auditor · skill-auditor

Score any skill, repo, or agent answer 0–8 with file:line evidence, before you trust it.

4 SKILLS

token-efficiency

context-optimization · context-thrift · lean-replies · output-sandbox

Make long-running agents spend attention deliberately.

1 SKILL

rag

composable-wiki-rag

Ground an answer in sources an operator can inspect.

1 SKILL

memory

layered-agent-memory

Keep durable context explicit, linked, and removable.

7 SKILLS

apple

app-integration · app-planning-per-platform · apple-fitness · healthkit · metal-shaders · widgets · xcode-mcp

Useful local actions with a visible permission surface.

λ5 SKILLS

swift

embedded-swift · react-native-to-swift · repo-to-swift-native · swift-on-android · swift-vapor

Swift beyond the app: embedded, Android, Vapor, ports.

3 SKILLS

workflow

grill-me · superpowers-tdd · superpowers-writing-plans

Turn a vague task into inspectable steps.

15 SKILLS

tools-services

cli-anything-browser · cli-anything-chromadb · cli-anything-exa · cli-anything-mermaid · cli-anything-obsidian · cli-anything-ollama · cli-anything-zotero · github-org-automation · use-cli-anything · use-crawl4ai · use-gortex · use-hermes · use-n8n · use-ntfy · use-tailscale

Connect agents to services without hiding the blast radius.

2 SKILLS

web

seo-white-hat · web3-wallet-integration

SEO without tricks, and wallet integration on the web.

1 SKILL

iot

led-grow-controller

Small interfaces deserve the same scrutiny as large ones.

read-only by design

The auditors inspect manifests, instructions, permissions, and scripts. They never execute the target skill during review; the score is a review signal, not a promise of safety.