pulser-eval

name	pulser-eval
description	Run cross-model evals on the Pulser for Odoo platform. Requires explicit resource_config documentation to ensure comparable results.
disable-model-invocation	true
user-invocable	true
allowed-tools	Bash(python3 scripts/run_eval.py)
inputs	[{"dataset":"String - path to the eval dataset."},{"resource_config":"String (e.g., 'high-headroom-v1') - The ACA resource footprint for this eval."},{"odoo_version":"String (e.g., '18.0')"}]

pulser-eval

This skill executes high-fidelity evaluations against Odoo 18 CE environments.

Eval Infrastructure Rule (Anthropic Engineering Blog)

Eval scores are not comparable across different infrastructure configurations. A 3x ceiling over per-task baseline specs is required to eliminate OOM-skew.

Registration of `resource_config` is mandatory.

Usage

• Trigger: "run evals on the AP flow"
• Required metadata: `resource_config`