eval-magic 0.5.0

One-stop CLI for running skill evals — measure whether an agent skill actually shifts behavior.
Documentation
{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "$id": "https://slow-powers.dev/schemas/judge-tasks.schema.json",
  "title": "Judge Tasks",
  "description": "Output of evals:grade (emit mode). The list of LLM judge tasks the orchestrator dispatches, plus the skill-invocation meta-checks. Lives at <workspace>/iteration-N/judge-tasks.json. The full prompt is written to dispatch_prompt_path and is NOT inlined here.",
  "type": "object",
  "required": [
    "generated",
    "total_tasks",
    "meta_tasks_injected",
    "skipped_transcript_checks",
    "tasks"
  ],
  "additionalProperties": false,
  "properties": {
    "generated": { "type": "string", "description": "ISO timestamp" },
    "total_tasks": { "type": "integer" },
    "meta_tasks_injected": { "type": "integer" },
    "skipped_transcript_checks": { "type": "integer" },
    "tasks": {
      "type": "array",
      "items": { "$ref": "#/definitions/task" }
    }
  },
  "definitions": {
    "task": {
      "type": "object",
      "required": [
        "eval_id",
        "condition",
        "assertion_id",
        "rubric",
        "model",
        "is_meta",
        "run_record_path",
        "outputs_dir",
        "response_path",
        "dispatch_prompt_path"
      ],
      "additionalProperties": false,
      "properties": {
        "eval_id": { "type": "string" },
        "condition": { "type": "string" },
        "run_index": {
          "type": "integer",
          "minimum": 1,
          "description": "1-based run index within a multi-run (eval, condition) cell; absent for single-run cells."
        },
        "assertion_id": { "type": "string" },
        "rubric": { "type": "string" },
        "model": {
          "type": ["string", "null"],
          "description": "Resolved judge model for this task, after assertion-level override and run-level default precedence; null means use the harness default."
        },
        "is_meta": { "type": "boolean" },
        "run_record_path": { "type": "string" },
        "outputs_dir": { "type": "string" },
        "response_path": { "type": "string" },
        "dispatch_prompt_path": {
          "type": "string",
          "description": "Absolute path to the file holding the full judge prompt."
        }
      }
    }
  }
}