{
  "context": "Tests whether the agent produces a migration runbook that accurately reflects the specific operational instructions in the migrate-to-tessl skill. The runbook must cover the non-obvious preflight requirements (three specific checks: CLI, login, required skills), the correct timing for Tessl project creation (before Phase 6 evals, not later), the correct default values, and key pitfalls from the lessons-learned reference.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "CLI version check",
      "description": "migration-runbook.md mentions running `tessl --version` (or equivalent) as a preflight check, and specifies that migration should STOP if the CLI is missing",
      "max_score": 8
    },
    {
      "name": "Login check",
      "description": "migration-runbook.md mentions running `tessl whoami` (or equivalent) as a preflight check to verify the user is logged in",
      "max_score": 8
    },
    {
      "name": "Required skills check",
      "description": "migration-runbook.md lists checking that required Tessl skills are available, and names at least two of: tessl__tile-creator, tessl__setup-skill-performance, tessl__optimize-skill-performance-and-instructions, tessl__compare-skill-model-performance",
      "max_score": 10
    },
    {
      "name": "Project creation timing",
      "description": "migration-runbook.md specifies that `tessl project create` must be run BEFORE the eval phase (Phase 6), not after publish — and does NOT suggest creating the project only when eval fails",
      "max_score": 12
    },
    {
      "name": "Eval caching pitfall",
      "description": "migration-runbook.md mentions that eval results are cached and that `--force` must be passed to `tessl eval run` to get fresh results after a content change",
      "max_score": 10
    },
    {
      "name": "Activation vs content eval",
      "description": "migration-runbook.md distinguishes between activation eval (tests routing, uses --solver=activation) and content eval (tests skill uplift), noting that both should be run",
      "max_score": 10
    },
    {
      "name": "Default visibility private",
      "description": "migration-runbook.md specifies that the default tile visibility is `private`",
      "max_score": 6
    },
    {
      "name": "Default version 0.1.0",
      "description": "migration-runbook.md specifies that the default initial version is `0.1.0`",
      "max_score": 6
    },
    {
      "name": "Default eval agent",
      "description": "migration-runbook.md specifies `claude:claude-sonnet-4-6` as the default eval agent",
      "max_score": 8
    },
    {
      "name": "Dry-run before publish",
      "description": "migration-runbook.md mentions running `tessl tile publish --dry-run` before the real publish",
      "max_score": 7
    },
    {
      "name": "Minor bump for final publish",
      "description": "migration-runbook.md specifies using `--bump minor` for the final publish after multi-model verification (Phase 7), as opposed to `--bump patch`",
      "max_score": 8
    },
    {
      "name": "Stop conditions listed",
      "description": "migration-runbook.md lists at least three stop conditions (e.g. missing CLI, not logged in, missing required skills, source cannot be parsed, project cannot be created, scenario gen stuck >30 min, three consecutive eval runs below 80%)",
      "max_score": 7
    }
  ]
}