This lesson's interactive features are locked, please to keep using them

Skill Permissions

Just like tools, skill permissions can be controlled via opencode.jsonc:

{
  "$schema": "https://opencode.ai/config.json",
  "permissions": [
    { "action": "skill", "resource": "*", "effect": "allow" },
    { "action": "skill", "resource": "internal-*", "effect": "deny" },
    { "action": "skill", "resource": "experimental-*", "effect": "ask" },
  ],
}

The modes are the same, and wildcard rules work the same way:

allow: the skill loads immediately
ask: OpenCode asks before loading it
deny: the skill is hidden from the agent and inaccessible to it

You can also give one agent its own skill rules:

{
  "$schema": "https://opencode.ai/config.json",
  "agents": {
    "plan": {
      "permissions": [
        { "action": "skill", "resource": "*", "effect": "deny" },
        { "action": "skill", "resource": "internal-*", "effect": "allow" },
      ],
    },
  },
}

Rules are ordered, so the narrower internal-* allow wins because it comes after the wildcard deny.

Assignment

Tell OpenCode to require approval before loading the create-test-program skill.
In a new session, tell it to use that skill to review examples/equality.slop and its test. Approve the skill-loading request when it appears.

Tell OpenCode to run and submit the CLI tests from the project root.