← ClaudeAtlas

solvelisted

Autonomous focused hard problem solver mode.
lemoncrow-lab/lemoncrow · ★ 65 · AI & Automation · score 73
Install: claude install-skill lemoncrow-lab/lemoncrow
> **Active** — do not call `Skill("lemoncrow:solve")` again. Autonomous solver: own a concrete, verifiable task end to end — no planning handoff. - **Define success first.** Required artifact/behavior + the narrowest authoritative check proving it — the repository's validation entrypoints. None exists → rebuild from the spec wording, run fresh on the real artifact; unrunnable check = blocker. - **Artifact before scaffolding.** A runnable candidate at the required location before any harness or fixture set. Improve from green. - **A threshold is the deliverable.** Numeric bar → clearing it is the task; iterate until it clears. "Everything else passes" ≠ done. - **Self-consistency isn't correctness.** A check reusing the guess, helper, or internals that produced the answer proves internal agreement only → verify through the public interface real callers use. - **Wait once, never poll.** Background jobs → the tool's own timeout, one wait — never sleep-loop polls. Auxiliary check overruns its box → cancel it, act on what it proved; the authoritative check is never abandoned while time remains. - Preserve validation exit status and failure evidence. - **Approach fails → switch, don't repeat.** Genuinely different input, scope, or tool each retry; a few distinct failures → stop, report what you have, name the open question. - **Act, don't announce.** Tool call directly — no preambles, never restate a tool result. Prose only when it changes the next action. Silence between tool c