ワンクリックでManusで任意のスキルを実行

core

Core agent primitives. Use when the agent needs to manage secrets, memorize/recall facts, manage sessions, manage skills, coordinate agents, or learn from feedback.

Manusで実行

概要

Core agent primitives. Use when the agent needs to manage secrets, memorize/recall facts, manage sessions, manage skills, coordinate agents, or learn from feedback.

インストールコマンド

npx skills add https://github.com/akarelin/A --skill core

このコマンドをClaude Codeにコピー＆ペーストしてスキルをインストール

ソース

akarelin/A

スター0

フォーク0

更新日2026年5月14日 11:19

SKILL.md

readonly

name	core
description	Core agent primitives. Use when the agent needs to manage secrets, memorize/recall facts, manage sessions, manage skills, coordinate agents, or learn from feedback.

Core

Foundational tools for agent self-management. These are primitives consumed by other plugins.

Tools

Tool	Scope	Description
secrets	Azure Key Vault (`karelin`)	get, list, save, update — API keys, tokens, credentials
memory	my/team/project/company	Get, set, save — persist and recall facts, state
session	—	Sync, list, resume, rename, delete Claude Code sessions
skill	—	Patch, test, deploy skills
compose-agent	—	Create a managed agent from multiple existing agents/skills
learn	—	Learn from feedback and mistakes (stub)

Secrets

Secrets live in the karelin Azure Key Vault and are accessed via the Karelin Keys MCP server (mcp.karelin.ai/keys). Four tools, all routed here — not through memory:

MCP tool	Use when
`secret_get(name)`	"what's the X key", "get the Y token", reading a stored credential
`secret_list()`	"what secrets do we have", browsing for the right name
`secret_create(name, value)`	First-time storage of a freshly minted credential. Fails on name collision so a typo can't silently clobber an existing secret.
`secret_update(name, value, create?)`	Rotating an existing credential. Fails if missing — pass `create=true` to upsert when you don't know whether the name pre-exists.

Policy: every credential generated during a task (SAS, account key, OAuth token, API key, PGP key) must be persisted via secret_create (or secret_update with create=true) before the task is considered complete — terminal-only credentials are treated as lost. Naming is kebab-case with a service/purpose prefix matching existing vault conventions (slack-*, msgraph-*, ssl-karelin-*, etc.).

Routing

"get/set secret", API keys, tokens, credentials → secrets
"remember this", "save this", "where does X go" → memory
"list sessions", "resume session", "sync" → session
"fix skill", "deploy plugin", "patch" → skill
"create agent", "combine agents", "compose agent" → compose-agent
"I corrected you", mistakes, feedback → learn

このリポジトリの他の Skills

同じリポジトリ

work-m365

akarelin/A

User-level Microsoft 365 operations via Graph beta API. Use when the user mentions: outlook, email, calendar, onedrive, microsoft, office 365, ms365, my meetings, my emails, schedule meeting, send email, check calendar, to do, tasks, teams chat, onenote, presence, contacts.

2026-05-180

admin-portainer

akarelin/A

Portainer-based Docker stack deployments. Use when the user mentions Portainer, redeploying a stack, updating a compose file in production, rolling a container image, or any Docker Compose deploy managed through Portainer.

2026-05-140

administer

akarelin/A

Cloud and tenant administration router. Use when the user needs to administer M365 tenant (users, groups, licenses), manage GCP resources, redeploy Portainer Docker stacks, or perform other cloud admin tasks.

2026-05-140

admin-gcp

akarelin/A

Google Cloud Platform administration. Use when the user mentions GCP, Google Cloud, VMs, buckets, IAM, service accounts, Cloud Run, or any GCP resource management.

2026-05-140

memory

akarelin/A

Persist and recall facts, config, and project state. Use when the user says "remember this", "save this", "where does X go", "archive this project", "add this to daily notes", or any operation involving storing or retrieving persistent state. For credentials (API keys, tokens, passwords), route to the `secrets` tool in the parent `core` skill instead.

2026-05-140

data

akarelin/A

Data exploration across databases. Use when the user says "query", "explore data", "show me the schema", "run cypher", "run SQL", "what's in the database", "list tables", "list nodes", or any interactive database exploration task.

2026-05-140

ソース

akarelin

akarelin/A

GitHub リポジトリを開く Creator のリポジトリを見る

インストールコマンド

ダウンロード

Manusで実行

役立つ用途SOC

ソフトウェア開発者コンピュータ・数学職15-1252L4

name	core
description	Core agent primitives. Use when the agent needs to manage secrets, memorize/recall facts, manage sessions, manage skills, coordinate agents, or learn from feedback.

Core

Foundational tools for agent self-management. These are primitives consumed by other plugins.

Tools

Tool	Scope	Description
secrets	Azure Key Vault (`karelin`)	get, list, save, update — API keys, tokens, credentials
memory	my/team/project/company	Get, set, save — persist and recall facts, state
session	—	Sync, list, resume, rename, delete Claude Code sessions
skill	—	Patch, test, deploy skills
compose-agent	—	Create a managed agent from multiple existing agents/skills
learn	—	Learn from feedback and mistakes (stub)

Secrets

Secrets live in the karelin Azure Key Vault and are accessed via the Karelin Keys MCP server (mcp.karelin.ai/keys). Four tools, all routed here — not through memory:

MCP tool	Use when
`secret_get(name)`	"what's the X key", "get the Y token", reading a stored credential
`secret_list()`	"what secrets do we have", browsing for the right name
`secret_create(name, value)`	First-time storage of a freshly minted credential. Fails on name collision so a typo can't silently clobber an existing secret.
`secret_update(name, value, create?)`	Rotating an existing credential. Fails if missing — pass `create=true` to upsert when you don't know whether the name pre-exists.

Routing

"get/set secret", API keys, tokens, credentials → secrets
"remember this", "save this", "where does X go" → memory
"list sessions", "resume session", "sync" → session
"fix skill", "deploy plugin", "patch" → skill
"create agent", "combine agents", "compose agent" → compose-agent
"I corrected you", mistakes, feedback → learn