Skip to content

Security: erfnzdeh/arvancloud-agent-skill

Security

SECURITY.md

Security

This skill drives a cloud account: DNS, servers, buckets, and TLS certificates. A leaked key or a prompt that writes one into the repo is the failure mode that matters.

Reporting

Do not open a public issue for:

  • a prompt or example that would put an API key, S3 secret, or acme.sh token into git
  • a way the skill tells an agent to echo ARVAN_KEY (or whatever the env is called) into a ticket, a log, or assets/config.example.json
  • a cert-deploy hook that would copy a private key to a host you did not name

Use a private vulnerability advisory. Redact keys. A leaked machine-user key is revoked in npanel under IAM, not by opening an issue.

In scope

  • Instructions that persist secrets inside the skill directory
  • Auth header mistakes that would send a key to the wrong host
  • Object Storage examples that confuse the machine-user key with S3 credentials
  • dns_arvan / acme.sh steps that leave a token in a world-readable file

Out of scope

  • ArvanCloud's own API, panel, or IAM. Report those to them.
  • A 403 from an under-permissioned key. That is IAM, not this repo.
  • Anyone using the skill against an account they own. That is the point.

Operator notes

assets/config.example.json is names only. Real state belongs in ~/.config/arvan/config.json, which is outside the skill. If a key has been in a chat or a screenshot, revoke it and make another machine user.

There aren't any published security advisories