Aee Review Skill

AEE Review

Diff two rich AEE assessments and emit the bounded queue a reviewer actually owes attention: claims whose propagated score moved by at least the materiality threshold, policy verdict flips, and added/removed claims.

User input

$ARGUMENTS

Accept previous=<path> and current=<path> (assessment JSON files, usually under .specify/extensions/aee/assessments/), optional materiality=<0..1> (default 0.05, the engine's validated hysteresis point), and optional limit=<n> top-k budget applied after sorting by movement size.

Prerequisites

1. Do not install the engine yourself. The runner uses an aee engine in the range >=1.6.0,<2 from the active Python; otherwise its first run installs applied-epistemic-engineering>=1.6.0,<2 into .specify/extensions/aee/.runtime (this needs network access once). If the runner refuses (exit 2) because that install is disabled (SPECKIT_AEE_AUTO_INSTALL=0) or failed, stop and relay its message, which names the manual python -m pip install "applied-epistemic-engineering>=1.6.0,<2" command.
2. Resolve both assessment paths inside the project root. Never follow symlinks or read outside the project root.

Execution

Run:

python .specify/extensions/aee/scripts/python/run_aee.py review --previous "<path>" --current "<path>"

The runner prints the review queue as JSON. Report the moved claims (largest first, with before/after scores and bands), any verdict changes, and added/removed claims. An empty queue is a real result: it means nothing moved materially between the two assessments — say so rather than re-reviewing everything.