Skills/agent-evals/assets/trigger-queries.copilot-usage-stats.json

[
  {
    "id": "pos-01",
    "query": "how many tokens has this project used so far?",
    "should_trigger": true,
    "split": "train"
  },
  {
    "id": "pos-02",
    "query": "what did this repo cost me in Copilot over the last month",
    "should_trigger": true,
    "split": "train"
  },
  {
    "id": "pos-03",
    "query": "which model has burned the most context in this workspace",
    "should_trigger": true,
    "split": "train",
    "note": "asks for attribution, not for advice on spending less"
  },
  {
    "id": "pos-04",
    "query": "how much of my input is cache reads versus actually fresh context",
    "should_trigger": true,
    "split": "train",
    "note": "the cache_read subset fact is the whole point of the Skill"
  },
  {
    "id": "pos-05",
    "query": "break my usage down by day for the last two weeks",
    "should_trigger": true,
    "split": "validation",
    "note": "never says token, cost, or Copilot"
  },
  {
    "id": "pos-06",
    "query": "am I spending more through the CLI or through the chat view on this repo",
    "should_trigger": true,
    "split": "validation",
    "note": "agent_name split across surfaces"
  },
  {
    "id": "pos-07",
    "query": "turn those tokens into actual dollars for me",
    "should_trigger": true,
    "split": "train",
    "note": "the AI-credit conversion; 1 credit = $0.01"
  },
  {
    "id": "pos-08",
    "query": "how many AI credits did this workspace eat this month",
    "should_trigger": true,
    "split": "validation"
  },
  {
    "id": "neg-01",
    "query": "give me some tips for reducing how many tokens I burn",
    "should_trigger": false,
    "split": "train",
    "note": "advice, not measurement — belongs to chronicle cost-tips"
  },
  {
    "id": "neg-02",
    "query": "write my standup for yesterday",
    "should_trigger": false,
    "split": "train",
    "note": "chronicle standup"
  },
  {
    "id": "neg-03",
    "query": "find the session where I fixed the compaction hook",
    "should_trigger": false,
    "split": "train",
    "note": "chronicle session search"
  },
  {
    "id": "neg-04",
    "query": "does my new skill actually get picked up reliably or was that one lucky run",
    "should_trigger": false,
    "split": "train",
    "note": "agent-evals trigger-rate work"
  },
  {
    "id": "neg-05",
    "query": "which model should I hand this subagent",
    "should_trigger": false,
    "split": "validation",
    "note": "subagent-dispatch model selection, not usage attribution"
  },
  {
    "id": "neg-06",
    "query": "this deployment has been running for 40 minutes, is it stuck or still working",
    "should_trigger": false,
    "split": "validation",
    "note": "long-running-job-monitor"
  },
  {
    "id": "neg-07",
    "query": "should I upgrade from Copilot Pro to Pro+ or Max",
    "should_trigger": false,
    "split": "validation",
    "note": "plan selection, not measurement of what was consumed"
  }
]