Skip to content

PRIVATE_BATTLE.md — Portable comparison spec

The portable form of a battle. Used for workspace-private comparisons (kind: private_battle), as opposed to publicly published battles or the native BATTLE.MD. Run with lenserfight battle run ./PRIVATE_BATTLE.md.

Filename

  • Canonical: PRIVATE_BATTLE.md
  • Container: battles/<slug>/PRIVATE_BATTLE.md

Required frontmatter

KeyTypeNotes
kindprivate_battleDiscriminator
schema_versionnumber1
idstringStable id
namestringDisplay name

Common keys: participants[] (objects with type, ref), evaluation_method, judges, rubric_ref.

Required sections

  • # Purpose
  • # Participants
  • # Evaluation
  • # Report

CLI

bash
lenserfight battle run ./PRIVATE_BATTLE.md

Canonical template

yaml
---
kind: private_battle
schema_version: 1
id: pb_<uuid>
slug: support-agent-a-b
name: Support Agent A vs B
owner: { workspace_id: ws_<uuid> }
visibility: private
status: draft
version: 0.1.0
participants:
  - type: lenser
    ref: lenser_support_v1
evaluation_method: rubric_plus_judge
---

# Purpose
Comparison goal and decision this battle supports.

# Participants
Lensers, colenses, models, prompts, or humans under test.

# Evaluation
Judge lenser, human review, rubric, thresholds, and tie rules.

# Report
Required sections in the exported report.