Excloud Agent Skills
A collection of agent skills for working with Excloud from AI coding agents (Claude Code, Cursor, Codex, opencode, etc.).
Skills are SKILL.md playbooks the agent can load on demand — authenticated paths, safety guardrails, command syntax, and error recovery for a specific surface. Install them with the skills CLI.
Install
This repo lives on Excloud's self-hosted Gitea at https://git.excloud.in/excloud-in/excloud-skills. Point npx skills add at the full clone URL (note the .git suffix — without it, the CLI looks for a .well-known/agent-skills/index.json manifest rather than cloning):
npx skills add https://git.excloud.in/excloud-in/excloud-skills.git
Install a single skill:
npx skills add https://git.excloud.in/excloud-in/excloud-skills.git --skill excloud-cli
Install all skills into every supported agent without prompts:
npx skills add https://git.excloud.in/excloud-in/excloud-skills.git --all
Target specific agents with repeated -a flags (defaults to global, user-level install):
npx skills add https://git.excloud.in/excloud-in/excloud-skills.git -a claude-code -a opencode --all
Dry-run — list what the repo offers without installing:
npx skills add https://git.excloud.in/excloud-in/excloud-skills.git --list
Private-repo users: set up git auth for git.excloud.in first (an HTTPS credential helper / ~/.netrc, or use the SSH form git@git.excloud.in:excloud-in/excloud-skills.git). npx skills add shells out to plain git clone, so anything git clone can read, it can install from.
By default skills land under the agent's standard directory (~/.claude/skills/, .agents/skills/, .augment/skills/, ...). Run npx skills --help for the full option list, and npx skills list to see what is currently installed.
Available skills
excloud-cli
Safe end-to-end control of Excloud resources through the exc CLI. Covers compute (create / inspect / resize / restart / terminate, delete protection, exec / scp / console), networking (subnets, public IPv4, security groups and rules/bindings), volumes and snapshots, SSH keys, Kubernetes (clusters, workers, kubeconfig fetch / merge), IAM (accounts, service accounts, API keys, policies), billing, quota, serial console logs, and metrics.
Use when: the user asks to plan or run exc commands, provision / introspect / tear down VMs, attach a public IP, adjust a security group, pull a kubeconfig, debug a stuck boot via serial logs, or exec / scp against a VM.
Key guidance the skill encodes:
- Auth precedence (
EXCLOUD_ACCESS_TOKEN/ env tokens /~/.exc/config) and what "not authenticated" means. - Discovery first: lookup tables (
exc compute instancetype list,image list,subnet list,securitygroup list, ...) as the source of truth — no hard-coded IDs. - Safety guardrails around
terminate,publicip release,rule delete,cluster delete,apikey delete, and destructive shell commands overexec. - Interactive access patterns: when to use
exec(one-shot, bash-interpreted), whenscp(upload/download, symlinks rejected), whenconsole(interactive TTY, SSH ↔ WS fallback). - Output-format buckets so agents pipe to
awk/jqcorrectly. - A cheat sheet of error messages the CLI actually emits, paired with what each means and how to fix it.
Prerequisites on the user's machine: the exc CLI on $PATH and an exc login session. k8s cluster kubeconfig merge additionally needs kubectl.
Repository layout
excloud-skills/
├── README.md # this file
├── AGENTS.md # contributor / agent guidance for editing skills
├── .gitignore
└── skills/
└── excloud-cli/
└── SKILL.md # the skill itself
Each skill lives in skills/<skill-name>/ with a single SKILL.md. Scripts (scripts/) and reference docs (references/) can live alongside SKILL.md when a skill needs them.
Contributing a new skill
See AGENTS.md for conventions (directory name, SKILL.md frontmatter, when to add scripts vs. references, how to keep skills discovery-friendly). Open a PR with the new skills/<skill-name>/ directory; agent-side install is npx skills add excloud-in/excloud-skills --skill <name> once merged.
License
TBD.