← ClaudeAtlas

verification-planninglisted

Plans how to prove a change works or broke nothing: frames the claim, designs an evidence path from the system itself, and requires every check to be able to fail.
mdrubelamin2/omc-slim · ★ 1 · AI & Automation · score 77
Install: claude install-skill mdrubelamin2/omc-slim
# Verification Planning ## Scope Use this skill proportionately. Small mechanical changes can follow ordinary project checks directly. For larger multi-phase work, let this skill establish the evidence path that later work follows. Invoked by name, it runs at any size. ## Who runs it, and with what This skill plans; other components execute. **Everything below is what the plan specifies, not what you personally perform**: the revert-and-re-run ritual, the runnable path, the closing interpretation. You write them into the plan and name their owner. Send the locating work to the `omc-slim:explorer` agent: which checks already exist, what they cover. The resulting edits go to a general-purpose writer briefed with the plan. Where the plan is one stage of a larger migration, the `omc-slim:deepwork` skill owns the sequencing and this owns each stage's failable check. And if you are already inside the `omc-slim:deepwork` skill, this *is* that stage's check. Do not re-enter the sequencing skill to plan one stage of the plan you are already running. The evidence path uses whatever this machine actually has. A test runner, a browser or devtools server, a database or API client, a coverage tool: any of them turns an assertion into an observation. They arrive from the project's `.claude/` and the user's `~/.claude/`, their names say nothing about their subject, and `ToolSearch` reaches them where tools are deferred. Read the descriptions, prefer a tool built for this stack over a ge