active-learning-experiment-planner

What problem does it solve?

It helps coding agents decide which learning experiment to run next using existing ee observation, uncertainty, and evaluation records, while preventing unsafe or unsupported closeout decisions.

Core Features & Use Cases

• Evidence-gated experiment planning from ee learn summaries, uncertainty reports, and evaluation/causal/economy/context bundles.
• Safety and data-integrity boundaries that forbid direct memory mutation and require explicit dry-run or user confirmation for follow-up actions.
• Stop/go refusal modes for empty observations, malformed inputs, degraded dependencies, underpowered samples, failed redaction, and unquarantined prompt-injection-like evidence.

Quick Start

Ask your agent: plan the next learning experiment for my workspace using my existing ee learning ledger data, and output a measurable dry-run experiment with follow-up ee commands.