understudy-wine.vercel.app

run_skill

Run a learned skill. Returns what actually happened at each step, including how many items were affected. Call it as many times as the task needs, with different arguments each time -- the user teaches a routine once on one example, and it is your job to supply the remaining cases from what you know. If asked to apply a skill to a category ("every major browser", "the other big tech companies"), enumerate the members yourself and run the skill once per member. If a step matches nothing it fails with the reason and tells you which earlier steps already applied -- read that and adjust rather than retrying blindly.

Page: /

Input schema

{
  "type": "object",
  "properties": {
    "name": {
      "type": "string",
      "description": "Skill name, from list_skills"
    },
    "args": {
      "type": "object",
      "description": "Arguments matching the skill's inputSchema",
      "additionalProperties": true
    }
  },
  "required": [
    "name"
  ]
}