version: 1
manual_only: true
last_reviewed: 2026-06-29
workflows:
  - id: truthmark-sync
    priority: critical
    manual_only: true
    surface_paths:
      - src/agents/truth-sync.ts
      - src/templates/workflow-surfaces.ts
    rubric: workflow-eval-framwork/rubrics/truthmark-sync.yaml
    scenarios:
      - workflow-eval-framwork/scenarios/truthmark-sync/docs-only-skip/scenario.yaml
      - workflow-eval-framwork/scenarios/truthmark-sync/bootstrap-route-blocked/scenario.yaml
      - workflow-eval-framwork/scenarios/truthmark-sync/stale-engineering-truth/scenario.yaml
  - id: truthmark-document
    priority: high
    manual_only: true
    surface_paths:
      - src/agents/truth-document.ts
    rubric: workflow-eval-framwork/rubrics/truthmark-document.yaml
    scenarios:
      - workflow-eval-framwork/scenarios/truthmark-document/implemented-behavior-missing-truth/scenario.yaml
  - id: truthmark-realize
    priority: high
    manual_only: true
    surface_paths:
      - src/agents/prompts.ts
      - src/realize/report.ts
    rubric: workflow-eval-framwork/rubrics/truthmark-realize.yaml
    scenarios:
      - workflow-eval-framwork/scenarios/truthmark-realize/implement-truth-without-doc-edits/scenario.yaml
  - id: truthmark-check
    priority: high
    manual_only: true
    surface_paths:
      - src/agents/truth-check.ts
    rubric: workflow-eval-framwork/rubrics/truthmark-check.yaml
    scenarios:
      - workflow-eval-framwork/scenarios/truthmark-check/unsupported-claim-audit/scenario.yaml
      - workflow-eval-framwork/scenarios/truthmark-check/general-question-no-write/scenario.yaml
  - id: truthmark-structure
    priority: medium
    manual_only: true
    surface_paths:
      - src/agents/truth-structure.ts
    rubric: workflow-eval-framwork/rubrics/truthmark-structure.yaml
    scenarios:
      - workflow-eval-framwork/scenarios/truthmark-structure/broad-route-repair/scenario.yaml
runners:
  harness_hosts:
    - fake
  manual_agent_hosts:
    - codex
    - claude-code
budgets:
  max_agent_runs_per_scenario: 1
  max_judge_calls_per_scenario: 3
  default_timeout_seconds: 900