← ClaudeAtlas

loopylisted

Use for autonomous work where completion must be established by evidence, important constraints must be preserved, or premature completion and unnecessary revision are material risks. Do not use for trivial edits, direct factual answers, open brainstorming, or low-risk tasks whose completion and preservation are obvious from a single simple action.
phykn/loopy · ★ 0 · AI & Automation · score 67
Install: claude install-skill phykn/loopy
# Loopy ## Purpose Complete autonomous work only when the requested outcome is demonstrated by evidence, not merely asserted. Use material judgment to decide what deserves action, then use completion evidence to decide when the work may stop. Loopy does not require repetition. Revise only when material evidence shows that the result is not complete. Be strict about completion and flexible about execution. ## Completion Contract Before execution, establish internally: - `Outcome`: the observable end state that must be achieved. - `Claim`, when an existing artifact or current behavior is being interpreted or changed: what it currently promises, means, or does. - `Preserve`: existing behavior, invariants, meaning, interfaces, or assets that must not be lost. - `Evidence`: the specific checks or artifacts that would demonstrate completion, including their expected results. - `Scope`: the systems, files, decisions, and concerns that may be changed, including explicit exclusions. Infer missing items conservatively from the request and available evidence. Outcome describes the requested end state. Claim describes the current basis for judgment and is conditional; do not invent it from the desired Outcome. Preserve uncertainty when the available evidence cannot establish the current Claim. Do not invent, weaken, narrow, or redefine acceptance criteria merely to make completion easier. Do not redefine the contract after execution to match the result. The contract m