Agent skill · software engineering · aiskillstore
subagent-driven-development
Sequential subagent execution with two-stage review gates for implementation plans. Use when executing multi-task plans in current session, when tasks need fresh subagent context to avoid pollution, when formal review cycles (spec compliance then code quality) are required between tasks, or when you need diff-based validation of each task before proceeding.
Why this skill is useful
Adds a comprehensive multi-task execution framework with specific templates and validation processes that enhance the AI's capability to manage complex development workflows.
What it needs
About 3k tokens when loaded. Last updated 2026-08-07. 405 stars on the source repository.
What this skill does
Subagent-Driven Development Dispatch fresh subagent per task with two-stage review after each. Core principle: Fresh subagent per task + two-stage review (spec then quality) + diff-based validation = high quality, fast iteration Quick Reference Phase Actor Action Exit Condition ------- ------- -------- ---------------- 1 Controller Read plan, extract tasks, capture BASESHA, create task list All tasks tracked with BASESHA 2 Subagent Implement (templates/implementer.md) Commit exists + tests pass + report includes HEADSHA 3 Subagent Spec review (templates/spec-reviewer.md) Reviewer checked diff(BASE..HEAD) and returned ✅ or file:line issues 4 Subagent Quality review (templates/code-quality-reviewer.md) Reviewer approved diff or returned severity-tagged issues 5 Controller Update state, mark complete, loop to step 2 All tasks done 6 Subagent Final review of entire implementation Approved 7 Controller Use finishing-a-development-branch Branch complete Task Contract Every task must satisfy these constraints: Scope: One task = one commit (or tight commit stack) No drive-by refactors; no formatting churn No unrelated file changes Verification: All tests pass (run full suite or affected subset) Lint/typecheck pass (if applicable) Commands actually executed (not assumed) Report must include: BASESHA (from controller) HEADSHA (after commit) Commands run (exact, copy-pasteable) Test output summary (pass/fail + counts) Files changed (list) Scope confirmation ("No unrelated changes") State Tracking Controller maintains per-task record: Update state after each phase transition. Process Setup (Once) 1. Read plan file once 2. Extract all tasks with full text and context 3. Capture current HEAD as BASESHA for first task 4. Create task list with all tasks (status: pending) Per Task Loop 1. Record BASESHA for this task (current HEAD) 2. Dispatch implementer with full task text + context + BASESHA 3. If questions: Answer completely, re-dispatch with answers 4. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aiskillstore/subagent-driven-development