A content platform builds a multi-agent pipeline where Agent A writes an article draft, Agent B edits it, and Agent C fact-checks it. The agents are chained: A's output is B's input, B's output is C's input. In production, they observe that errors in Agent A's output propagate and compound through the chain — B edits with the assumption A's claims are correct, then C fact-checks B's version but misses A's original error because B's rephrasing obscures it. How should the pipeline architecture be redesigned to prevent error compounding?

Error compounding in sequential pipelines occurs when downstream agents operate only on transformed versions of upstream output, losing visibility into the original source of a claim. The fix is to pass both the original draft (A's output) and the edited version (B's output) to Agent C — preserving the original formulation of claims even as editing has changed their wording. C can then fact-check the underlying assertions regardless of stylistic transformation. Option A (parallel execution) eliminates the sequential dependency but loses the benefit of having B's editing improve A's draft before C's fact check. Option C (validate A before editing) is excellent defense but doesn't help when A's error is subtle and passes initial validation — it catches obvious errors but not sophisticated factual mistakes. Option D (combining B and C) reduces the multi-agent benefit by making one agent responsible for both creative editing (which benefits from focus) and analytical fact-checking (which benefits from separate focus).