orchestrating-development-teamlisted

# Orchestrating the development team Coordinate the plugin's skills and agents (see [agents/](../../agents/)) through a fixed pipeline so multi-part work stays coherent, verifiable, and free of conflicting concurrent edits. This skill is the router for cross-cutting requests — it delegates depth to the other skills rather than duplicating their instructions. This skill itself follows [enforcing-safety-baseline](../enforcing-safety-baseline/SKILL.md). **Routing vs. execution — do not conflate the two.** Whether to *invoke* this skill is a routing decision: an explicit full-team request always invokes it, full stop, regardless of task size. Whether to *right-size* the pipeline once invoked (e.g. skip straight to `implementing-features` for a genuinely small change) is an internal execution decision this skill makes for itself, after it has already loaded — never a reason for the top-level router to skip loading it in the first place. ## Pipeline ``` DISCOVER → ARCHITECT → PLAN → IMPLEMENT → DATABASE → TEST → REVIEW → FIX → VERIFY → RELEASE → DEPLOY-PLAN → REPORT ``` Every stage is conditional on evidence — never run mechanically. See `references/orchestration-budget.md` for the stage policies (when REVIEW's security leg is mandatory, the test/database/architecture/release-deploy boundaries) referenced below. | Stage | Delegates to | Produces | Runs when | |---|---|---|---| | DISCOVER | `analyzing-codebase`, optionally `ai-dev-team:repository-explorer` for a large/unfamili