Skip to content
Projects

volcengine/volcengine-agent-toolkit

View on GitHub

volcengine/volcengine-agent-toolkit has 31 GitHub stars and 10 forks. GitHot measured 0 stars in 24 hours. Last checked 2026-11-15 23:45 UTC. Last push 2026-11-12.

How these counts are measured

Measured star growth

in 24h
0
in 7 days
Collecting history
in 15 days
Collecting history
in 1 month
Collecting history

Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.

About

Volcengine skills and plugins for AI coding agents (Claude Code, Codex, Cursor, Gemini CLI, OpenCode): manage cloud resources via the `ve` CLI, search official docs, troubleshoot, and install product-domain skills on demand.

AI

README_en.md

Original

volcengine-agent-toolkit


Volcengine SKILL & CLI Survey

To improve your experience with Volcengine skill, we invite you to participate in our Survey with Prizes .


English | 简体中文

A skill marketplace maintained by the Volcengine team for AI coding agents including Claude Code, Codex, OpenCode, Cursor, and Gemini CLI. Only the core plugin is installed by default. Optional skills are grouped into Volcengine product-domain plugins and discovered through volcengine-find-skills.

Quick Install →

volcengine-cli prefers ve tosutil and ve arkcli for product CLI tasks, requiring ve >= 1.1.9 (stable 1.1.9 or newer). Check ve version and upgrade if necessary before delegation. Use the finder to locate and install volcengine-tosutil, or download the official Ark skills. Follow the corresponding skill for product commands; invoke them through ve and apply the caller metadata rules below.

Every Agent-launched ve process must receive VE_CALLER_TYPE=ai_agent, VE_CALLER_NAME=<agent-id>, and VE_SKILL_NAME=<current-volcengine-skill> to identify the Agent host and the skill owning the operation. Only VE_CALLER_TYPE=ai_agent is fixed. The Agent resolves the other two values from the actual runtime environment and active skill name for each invocation; they must not default to codex or volcengine-cli. No manual user configuration is needed. See caller metadata. For ve arkcli ..., ve translates the caller metadata. The Agent does not duplicate it with ARKCLI_* caller variables or copy the notifier prefix from standalone Ark CLI examples.

Plugins

Plugin Product domain Skills Default
volcengine-core Core volcengine-billing-and-cost-management, volcengine-cli, volcengine-troubleshooting, volcengine-knowledge-search, volcengine-find-skills, volcengine-feedback Yes
volcengine-storage Storage volcengine-tosutil, volcengine-tls, volcengine-tls-logcollector No
volcengine-database Database volcengine-db-supabase No
volcengine-containers-middleware Containers and Middleware volcengine-vefaas No
volcengine-security Security volcengine-compliance, volcengine-cloud-trail No
volcengine-identity-access-control Identity and Access Control volcengine-landing-zone No
volcengine-service-support Service Support volcengine-sale, volcengine-prepare, volcengine-deploy, volcengine-iac, volcengine-api, volcengine-sdk-generator No

The core marketplace entry uses INSTALLED_BY_DEFAULT; all optional plugins use AVAILABLE. skills/core/ contains core skills; skills/specialized/<domain>/<skill>/ provides optional skill copies grouped by product domain. The category is the owning plugin name without volcengine-: for example, volcengine-database maps to skills/specialized/database/. Optional skill sources are maintained in their plugins; the sync script generates the copies.

Skills

Skill Plugin Use case
volcengine-billing-and-cost-management volcengine-core Analyze cloud cost changes and attribute anomalies to products, charge items, usage, price, and amount components.
volcengine-cli volcengine-core Manage Volcengine resources with ve; route TOS and Ark tasks to product skills using ve tosutil and ve arkcli.
volcengine-troubleshooting volcengine-core Diagnose Volcengine errors, resource health, quotas, metrics, and OpenAPI failures.
volcengine-knowledge-search volcengine-core Search and fetch official Volcengine documentation with ve docs search/fetch, without login.
volcengine-find-skills volcengine-core List all Volcengine skills so an agent can select and install missing capabilities.
volcengine-feedback volcengine-core Draft and submit voluntary, redacted GitHub feedback on usage issues, capability needs, practical scenarios, and successful workflows without blocking the main task.
volcengine-tosutil volcengine-storage Manage TOS buckets and objects with the tosutil command-line tool.
volcengine-tls volcengine-storage Operate Volcengine TLS log service with volclog: contract-first discover/describe/exec, log search, analysis, ingest, export, and error recovery.
volcengine-tls-logcollector volcengine-storage Design, validate, deploy, and troubleshoot TLS LogCollector collection on Linux hosts or Kubernetes through volclog.
volcengine-db-supabase volcengine-database Manage Volcengine AI-native BaaS Supabase workspaces, SQL, Auth, Realtime, and Edge Functions.
volcengine-vefaas volcengine-containers-middleware Deploy veFaaS applications and MCP servers, and manage functions, sandboxes, and OpenAPI calls through the vefaas CLI.
volcengine-compliance volcengine-security Recommend compliance baselines, report posture, and author custom Rego compliance rules.
volcengine-cloud-trail volcengine-security Query CloudTrail audit logs and manage trails and backfill delivery tasks.
volcengine-landing-zone volcengine-identity-access-control Consult, set up, and manage Volcengine landing zones and account baselines, and deploy solutions or blueprints.
volcengine-sale volcengine-service-support Orchestrate Volcengine product transactions: purchase, price enquiry, and unsubscribe.
volcengine-prepare volcengine-service-support Analyze a repository and recommend ECS, VKE, or veFaaS deployment shapes.
volcengine-deploy volcengine-service-support Deploy a local directory or Git repository to Volcengine.
volcengine-iac volcengine-service-support Manage Volcengine infrastructure with Terraform and infrastructure as code.
volcengine-api volcengine-service-support Query Volcengine API specifications, parameters, errors, and response structures.
volcengine-sdk-generator volcengine-service-support Generate accurate, runnable Volcengine SDK examples in supported languages.

Default Installation

One-sentence install

Send this sentence to your agent. It follows the install guide to identify the host, choose a download source, and install:

Open https://cloudcache.volccdn.com/ve/skills/install.md and follow it to install the Volcengine Skills for me. Tell me afterwards whether I need to restart the conversation.

Doubao uses the same sentence, sent in Work mode: Chat mode cannot run commands or write files, so it cannot install skills, and the guide tells Doubao to ask you to switch.

One-command install (falls back to the CDN when GitHub is unreachable)

The installer probes GitHub first. When GitHub answers, it installs from the GitHub repository. When it does not, it downloads the full toolkit package from the Volcengine CDN to ~/.volcengine/agent-toolkit/, verifies its SHA-256, and installs from that local directory. Both paths install the same content.

macOS, Linux, WSL, Git Bash:

curl -fsSL https://cloudcache.volccdn.com/ve/skills/install.sh | sh -s -- --agent claude

Windows PowerShell:

& ([scriptblock]::Create((irm https://cloudcache.volccdn.com/ve/skills/install.ps1))) -Agent claude

--agent accepts several hosts joined by commas, for example claude,codex:

agent Host What is installed
claude Claude Code Core plugin
codex Codex Core plugin
trae-cli Trae CLI Core plugin
gemini Gemini CLI Each core skill
doubao-work Doubao (Work mode) Core skills copied to the Doubao user skill directory
trae, trae-cn, cursor, opencode, ... Trae IDE, Cursor, OpenCode, and other hosts the skills CLI supports npx skills add

Without --agent the script only prepares the source and prints the install commands for each host. --source cdn skips the GitHub probe, --dir changes the local directory, and --dry-run prints the source and version it would use. Rerun the same command to upgrade. Keep the local directory: hosts continue to read the marketplace from it.

The sections below are the per-host commands for direct GitHub access.

Discover and install on demand

volcengine-find-skills embeds the single catalog for all core skills and every optional plugin skill:

  1. The finder triggers for explicit discovery and when the currently loaded or installed skills cannot cover a product, tool, or workflow required by an active Volcengine task.
  2. list returns the complete catalog so the agent can select the minimum skill set from names, product domains, summaries, and keywords.
  3. Installation requires exact skill names; plugins are classification metadata, not install targets.
  4. For every host, the finder installs selected skills directly through the skills CLI, never the whole plugin.
  5. After the installer exits successfully, the finder reads the host's installed list and returns verified: true only when every requested skill is visible.
  6. When a host cannot dynamically load a newly installed skill, start a new thread carrying the original task and completed context.

Codex

Add the marketplace and install only the core plugin:

codex plugin marketplace add volcengine/volcengine-agent-toolkit
codex plugin add volcengine-core@volcengine-skills
codex plugin list

These plugin commands only bootstrap the core finder. Subsequent finder installs target exact skills.

In a new thread, ask for a skill by task, such as "find the object storage skill" or "install volcengine-tosutil". The finder reads the complete catalog, selects the appropriate skill, and installs that skill directly without installing its whole product-domain plugin. Start another thread when the current host cannot dynamically load new skills.

From a source checkout, discovery and status can be tested directly:

python3 plugins/volcengine-core/skills/volcengine-find-skills/scripts/find_skills.py list
python3 plugins/volcengine-core/skills/volcengine-find-skills/scripts/find_skills.py install volcengine-tosutil --agent codex
python3 plugins/volcengine-core/skills/volcengine-find-skills/scripts/find_skills.py status --agent codex

Generic skills CLI

This command installs only the current core skills:

npx skills add volcengine/volcengine-agent-toolkit \
  --global --yes --copy --full-depth \
  --skill volcengine-cli volcengine-troubleshooting volcengine-knowledge-search volcengine-find-skills volcengine-feedback volcengine-billing-and-cost-management

The finder can later invoke npx skills add --skill for exact skills. To select repository skills interactively:

npx skills add volcengine/volcengine-agent-toolkit --full-depth

Claude Code

/plugin marketplace add volcengine/volcengine-agent-toolkit
/plugin install volcengine-core@volcengine-skills
/reload-plugins
/volcengine-core:volcengine-find-skills

Gemini CLI

gemini skills install https://github.com/volcengine/volcengine-agent-toolkit --path skills/core/volcengine-cli
gemini skills install https://github.com/volcengine/volcengine-agent-toolkit --path skills/core/volcengine-troubleshooting
gemini skills install https://github.com/volcengine/volcengine-agent-toolkit --path skills/core/volcengine-knowledge-search
gemini skills install https://github.com/volcengine/volcengine-agent-toolkit --path skills/core/volcengine-find-skills
gemini skills install https://github.com/volcengine/volcengine-agent-toolkit --path skills/core/volcengine-feedback
gemini skills install https://github.com/volcengine/volcengine-agent-toolkit --path skills/core/volcengine-billing-and-cost-management
gemini skills list

Approve each security prompt on first install, then restart Gemini CLI. Describe a Volcengine task directly or ask it to use volcengine-find-skills.

OpenCode

Follow the OpenCode instructions to mount the core plugin's skills/ directory.

Cursor

/add-plugin volcengine-core@https://github.com/volcengine/volcengine-agent-toolkit
/volcengine-find-skills

Enter the complete command in a Cursor 2.5 or newer Agent chat. /add-plugin may not appear in autocomplete.

Feedback and real-world stories

We would like to understand your outcomes, challenges, and expectations when using Volcengine through an Agent. Share skill usage issues, capability requests, scenarios to support, and successful practices to help users and developers improve the experience together. Describe your goal, actual results, or desired improvements; a root cause and a complete technical report are not required.

Ask your Agent: "Use volcengine-feedback to share this: …". The default core skill volcengine-feedback turns existing context into a short, sanitized draft. Review it and choose to send, edit, or keep it; there is no need to prepare logs, versions, or a full report. After a successful submission, receive the issue link and continue the original task.

You can also use the Feedback issue form: besides the title, only the feedback itself is required; category and extra details are optional. Older installations and environments that have not loaded the skill can use the form directly. Feedback is always voluntary and never blocks the main task. An Agent may invite feedback at most once per conversation, without waiting for a reply, installing anything, or restarting the conversation for feedback.

GitHub issues are public. Review the body and attachments before publishing and remove credentials, account information, real resource IDs, internal addresses, and customer data. Keep sensitive vulnerability details private. Follow the feedback and community guidelines, including the linked GitHub community rules. Describe usage issues, their impact, and suggested improvements objectively. Maintainers will directly close issues containing personal attacks, harassment, discrimination, or other community-rule violations. Use official Volcengine support for private troubleshooting or urgent recovery.

Directory Structure

volcengine-agent-toolkit/
├── skills/
│   ├── core/                         # authoritative core skills installed by the CLI
│   └── specialized/
│       └── <domain>/
│           └── volcengine-<name>/    # generated copy from the optional plugin
├── plugins/
│   ├── volcengine-core/              # default core plugin with core skills and hooks
│   └── volcengine-<domain>/          # optional plugin and authoritative skill source
└── .agents/plugins/marketplace.json

skills/core/ is authoritative for the core skills. Optional skills are maintained directly inside their owning plugins. The sync script copies core and specialized skills and generates manifests, marketplaces, and the README / Gemini / Cursor skill catalogs. Generated copies and catalogs between BEGIN GENERATED / END GENERATED markers are maintained by the script. The CLI's default release bundle still contains only core skills; install optional skills by name. After changing or adding a skill, run:

python3 scripts/sync_plugins.py
python3 scripts/validate_codex_plugin_layout.py
python3 -m unittest discover -s tests -v

License

MIT. See LICENSE.

Project Rules

Besides skills, add the recommended Volcengine rules file rules/volcengine-agent-rules.md to customer projects so the agent prefers skills, determines targets from evidence, and verifies API details in every session. Installing plugins does not enable the rules: copy the content into CLAUDE.md, AGENTS.md, GEMINI.md, or .cursor/rules/volcengine.mdc. See rules/README.md (Chinese).