## Definition A **sequential pipeline** composes multiple specialised agents in a fixed order, each handing off to the next. The most common shape: **Plan → Build → Verify**. ## Canonical Shape ``` [Human] │ "Implement openspec/changes/<slug>/proposal.md" ▼ [Architect] → produces plan.md │ ▼ [Builder] → implements per plan, runs tests │ ▼ [Reviewer] → diffs vs spec, flags gaps │ ▼ [Human] → approve / push back ``` ## Why It Works - Each phase has **one job** and one output format. - The reviewer reads the diff cold — independent of the builder's reasoning. - The human reviews the *reviewer's verdict* first, not the raw code. ## Operational Use In Claude Code, the parent agent (you, in the CLI) delegates via sub-agent invocations: ``` Coordinator: → Use Agent(architect) to produce a plan. → Read the plan. → Use Agent(builder) with the plan and the spec. → Use Agent(reviewer) on the resulting diff. → Surface the reviewer's verdict. ``` ## When to Reach For It - High-stakes changes where reversibility is low. - Cross-cutting work where the *plan* is the most valuable artifact. - Anything you'd want a teammate to review before you merge. ## When NOT to - Trivial changes (one-line bug fix). The pipeline overhead exceeds the value. - Throwaway experiments. - Tasks with no spec to verify against — see [[Spec-Driven Development]]. ## Cost Roughly 3–4× a single-agent run for a Plan → Build → Critic pipeline. Justified by the issues caught before human review — most teams report the pipeline pays back within a single saved revision cycle. ## Related - [[Specialized Agent]] - [[Builder-Critic Pattern]] - [[Parallel Agents via Worktrees]] - [[Spec-as-Prompt Pattern]]