Skip to content

Security: Hamza-Abili/opencode-agent-plugins

Security

SECURITY.md

Security

Loading third-party packages into an agent that runs tools is a security decision. This plugin's posture is the product.

Model

  • Skills are inert data (markdown + frontmatter). Loading a skill never executes code, but skill content enters the agent's context — treat it as untrusted prompt material, like any file an agent reads.
  • MCP servers are executable code (local processes or remote endpoints). They are always registered disabled; enabling one (/ap enable or a pasted block on v1) is an explicit, per-server decision. /ap enable and /ap disable persist; enabledMcp is only a per-session seed.
  • Containment (realpath, after symlink resolution):
    • plugin.json must resolve inside the plugin root, or the plugin is rejected.
    • skills/, each SKILL.md and mcp.json must resolve inside the plugin root, or that component/skill is skipped with a reported warning.
    • ./ MCP commands must exist as regular files inside the plugin root and are emitted as absolute post-containment paths, so the spawn cwd cannot redirect which executable runs. ../ and out-of-root absolute paths are rejected; bare commands resolve through PATH like any local tool.
    • cwd must exist and resolve inside the plugin root (plugin-relative and ${PLUGIN_ROOT} forms) or inside the plugin data directory (${PLUGIN_DATA} forms, created when missing); .. and escaping symlinks are rejected. The emitted value is always a realpath.
    • env may not set PLUGIN_ROOT or PLUGIN_DATA; the client sets them last, so a plugin can never re-point another server's runtime.
    • ${PLUGIN_ROOT}/${PLUGIN_DATA} are expanded in args/env/cwd only — never in command.
  • Remote MCP URLs must use https; plain http is allowed on loopback only. Embedded credentials, query strings and fragments are rejected.
  • Marketplace installs run only on explicit /ap install, accept https/ssh git URLs without credentials/query/fragment, reject .. traversal and marketplace-root escapes, validate the target as an Agent Plugin before completing, install through a staging path (never a half-installed directory), pin the resolved commit, and record provenance. uninstall only removes directories inside the managed store.
  • No silent configuration mutation. The v1 adapter only manages symlinks it created (state in $XDG_DATA_HOME/opencode-agent-plugins/v1-skills.json, recording both the source and the skills directory); it never edits opencode.json. The v2 adapter registers exclusively through plugin transforms.
  • User configuration always wins. Skill ids and MCP server names that already exist are skipped, never overwritten.
  • Zero runtime dependencies — no transitive supply chain.

Known limitations

  • Skill content is not sanitized; a malicious skill can try prompt injection against the agent (same exposure as any untrusted file an agent reads). Review skills before installing.
  • On v1, MCP opt-in is a manual config edit; the plugin prints the block but does not verify it after you add it. The kill switches (AGENT_PLUGINS_DISABLED, AGENT_PLUGINS_SKILLS=0) remove owned links from the recorded and current skills directories, but a directory that is not writable at that moment keeps its ownership record for a later retry.
  • Symlinked plugin directories are resolved before containment checks, so a plugin whose root itself is a symlink into another directory is loaded from its real location.
  • ./ commands must exist as regular files and cwd must exist at map time; both are emitted as realpaths. Local/npm plugin trees are assumed not to be mutated concurrently by an attacker; only marketplace git installs are copied into a managed store.
  • A marketplace entry's policy.authentication is informational: this client never auto-authenticates or auto-enables anything.

Reporting

Report vulnerabilities privately via GitHub's "Report a vulnerability" flow on the repository. Do not open a public issue for suspected vulnerabilities.

There aren't any published security advisories