independent-analysislisted

# independent-analysis A standing tool to counter the single-context biases of whatever model is driving the conversation — going fast, and confirming its own earlier assumptions. It runs the **`independent-analysis` workflow** (the 9th `@workflow-toolbox` composition, also bundled as the plugin workflow `workflow-toolbox:independent-analysis`): one analyst agent per lens, each *blind to this conversation*, finds forgotten angles; a synthesis step dedups them against what you already know; then `adversarialVerification` re-derives each survivor refute-first and reports a verdict. Because the agents can't see the chat, they can't inherit its blind spots — that is the whole point. ## When to reach for it - **Before committing to a non-trivial design, plan, or decision** — let independent agents try to break it first. - The user asks to **stress-test / red-team / poke holes / get a second opinion**, or says **"are you sure?", "what am I missing?", "did we forget anything?"**. - **Auditing docs, a spec, or a plan for accuracy** (it dogfooded exactly this: it audited its own catalog change and caught real stale counts a self-review missed). - You notice you're **moving fast or defending a prior conclusion** — the failure mode this exists to catch. **When NOT to:** a single quick fact lookup, a mechanical edit, or a task with no "could this be wrong?" surface — the fan-out isn't worth it. It **finds issues, it does not fix them**: apply confirmed findings yourself. ##