speclisted

# Spec Turn a request into a decision-ready, testable specification. Stop after the spec unless the user separately asks for execution. ## Goal Produce the smallest spec that makes the work unambiguous enough to execute and verify across software, infrastructure, research, operations, writing, data, or other domains. ## Input Use the user's request plus the current repository, artifact, environment, instructions, and authoritative sources that materially affect the task. Do not ask for information that can be discovered. Ask only when a missing decision would materially change the requested outcome and cannot be inferred safely. ## Output When `PROLOG_TMP_SPEC_CONTEXT` is set by an agent launcher, it is the authoritative scratch root for specification work. Write task specs under: ```text $PROLOG_TMP_SPEC_CONTEXT/spec/<task-slug>/SPEC.md ``` If `$PROLOG_TMP_SPEC_CONTEXT/context.prolog` exists and the task requests Prolog, RLM-style reasoning, symbolic verification, or a Prolog specification, update that file with the compact machine-readable requirements, observations, invariants, provenance, work state, and proof/test results needed by the run. Do not turn it into a transcript or source-code dump. Otherwise, when a writable temporary filesystem exists, write the spec to: ```text ${TMPDIR:-/tmp}/spec/<task-slug>/SPEC.md ``` Use this skill's `scripts/init-spec.py '<task>'` helper to create the canonical path and section scaffold. It prefers `PROLOG_TMP_SPEC_CONTEX