Choose or reconsider a semantic code boundary when ownership, representation, composition, or required laws are in question. Also use for distributed implementations of one law or invalidated decisions. Not routine work under an accepted architecture.
Explicit-only architectural review of a PR, subsystem, repository, or proposed design. Reconstruct obligations and interactions, challenge the decomposition, and use Universalist for bounded constructive alternatives. Return evidenced defects, risks,…
Explicit-only Elenctic adjudicates the authenticated gh user's prior review threads, then runs an exact-head PR review campaign for the current branch or an explicit PR, forking prepared file reviewers with at most 20 active at once, reconciling their…
Implement accepted goals, analyze counterexamples, or close out reviewed changes. Turn supported failure families into constructions that preserve required-valid behavior. Unqualified review or analysis is read-only.
Build AI-powered software with TypeSafe or review local text and Markdown with the typesafe CLI. Use when an app needs programmable common sense, typed judgments and probabilities from System One models such as Jev, brainstorming AI possibilities, or…
Use for deliberate Lean 4 work: proof repair, theorem development, verified programs, model/specification design, external-code models, state-machine or trace invariants, termination proofs, Std/mathlib theorem discovery, Lake/toolchain diagnosis, and…
Capture or recall evidence-backed execution learnings; evaluate capture at validation transitions, strategy pivots, footguns, retry loops, and material delivery or handoff. Also handle explicit learning-memory admission and supersession.
Implicitly invoke when implementation, debugging, review, or validation encounters a witnessed failed/no-effect attempt, benchmark or test regression, revert, repeated same-cluster retry, abandoned strategy, or asks what has already been tried. Project the…