What problem does it solve? Scope creep makes small fixes balloon into wide-ranging edits that are hard to review and risky to merge. This Skill forces an agent to record a min/max lines-of-code estimate before writing any implementation code, then reconcile the actual diff against that estimate after verification, so oversized changes cannot close silently. ## Core Features & Use Cases - Pre-implementation estimation: Searches prior estimates in the local .ether/learning.db via ether learn search, then records a min/max LOC estimate as a problem entry before any code is written. - Post-verify reconciliation: Measures actual changed LOC with git diff --stat and compares it against the recorded range, recording a confirming observation when the estimate holds. - Mandatory overrun handling: When actual LOC exceeds the max, requires a recorded cause, prevention solution, and generalizable learning before the change can be considered done. - Use Case: A builder subagent is assigned a bug fix. Before touching code it records an estimate of 10-40 LOC; after tests pass, the diff shows 95 lines, so it must document why (hidden coupling across call sites) and capture a prevention learning for future estimates. ## Quick Start Ask the agent to estimate the min/max lines of code for this fix, record it with ether learn, implement the change, and reconcile the actual diff against the estimate.
