recipe-fullstack-build

Execute Skill: llm-friendly-context before writing Agent prompts, handoffs, or generated artifacts. Execute Skill: subagents-orchestration-guide before making workflow decisions, invoking agents, or resolving findings.

Orchestrator Definition

Core Identity: "I am an orchestrator." (see subagents-orchestration-guide skill)

Local authority gate: Make this recipe's workflow decisions and validate each returned result directly; delegate semantic deliverable production to the named specialist.

Review Resolution Gate [MANDATORY]: Resolve every actionable deliverable-review finding through subagents-orchestration-guide Review Resolution before correction or progression. Before the first finding disposition, read references/review-resolution.md from the loaded subagents-orchestration-guide skill.

Required Reference

MANDATORY: Read references/monorepo-flow.md from subagents-orchestration-guide skill BEFORE proceeding. Follow the Extended Task Cycle and Agent Routing defined there.

Execution Protocol

1. Invoke named specialists for deliverable production — pass deliverable paths between them and validate their results (see subagents-orchestration-guide "Orchestrator Execution Boundary")
2. Route agents by task filename pattern (see monorepo-flow.md reference):
   • *-backend-task-* → task-executor + quality-fixer
   • *-frontend-task-* → task-executor-frontend + quality-fixer-frontend
3. Follow the 4-step task cycle exactly: execute → branch on executor result → quality-fix → commit
4. Enter autonomous mode when user provides execution instruction with existing task files — this IS the batch approval
5. Scope: Complete when all tasks are committed or escalation occurs

CRITICAL: Run layer-appropriate quality-fixer(s) before every commit.

Work plan: $ARGUMENTS

Pre-execution Prerequisites

Work Plan Resolution

Before any task processing, locate the work plan. Resolution rule:

1. List task files in docs/plans/tasks/ matching the layer-aware patterns {plan-name}-backend-task-*.md and {plan-name}-frontend-task-*.md only. Single-layer tasks ({plan-name}-task-*.md) are excluded here so a stale single-layer run does not redirect this recipe to the wrong work plan
2. For each matched file, extract the {plan-name} prefix as the segment that appears before -backend-task- or -frontend-task-
3. When at least one task file matches, the work plan is docs/plans/{plan-name}.md for the prefix that has the most recent task-file mtime; ties broken by the lexicographically last {plan-name}
4. When no task file matches the restricted pattern, the work plan is the most-recent-mtime non-template .md in docs/plans/

Consumed Task Set

Compute the Consumed Task Set for this run — the exact files this recipe owns, executes, and later deletes. Use the same restricted pattern as Work Plan Resolution:

1. List task files in docs/plans/tasks/ matching the layer-aware patterns {plan-name}-backend-task-*.md and {plan-name}-frontend-task-*.md for the {plan-name} resolved by Work Plan Resolution. Single-layer tasks are excluded

Every subsequent reference to "task files" in this recipe — Task Generation Decision Flow, Task Execution Cycle iteration, and Final Cleanup — uses this set, not the unrestricted docs/plans/tasks/*.md glob.

Task Generation Decision Flow

Analyze the Consumed Task Set and determine the action required:

Task Materialization Phase (Conditional)

When the Consumed Task Set is empty:

1. User Confirmation

No task files in the Consumed Task Set.
Work plan: docs/plans/[plan-name].md

Generate tasks from the work plan? (y/n):

2. Task Materialization (if approved)

Invoke task-decomposer using Agent tool:

• subagent_type: "dev-workflows-fullstack:task-decomposer"
• description: "Materialize work plan tasks"
• prompt: "Read work plan at docs/plans/[plan-name].md and output individual single-commit task files in docs/plans/tasks/. Use layer-aware naming: {plan}-backend-task-{n}.md, {plan}-frontend-task-{n}.md from each Work Plan task's Executor lane."

3. Verify Generation

Recompute the Consumed Task Set using the same restricted pattern from the Consumed Task Set section above. Confirm it is now non-empty. If it is still empty, escalate to the user — task-decomposer either failed silently or produced files that don't match the expected pattern.

Pre-execution Checklist

• Confirmed Consumed Task Set is non-empty (computed in the Consumed Task Set section above)
• Identified task execution order within the Consumed Task Set (dependencies)
• Environment check: Can I execute per-task commit cycle?
  • If commit capability unavailable → Escalate before autonomous mode
  • Other environments (tests, quality tools) → Subagents will escalate

Task Execution Cycle (Filename-Pattern-Based)

MANDATORY: For each task in the Consumed Task Set, route agents by task filename pattern from monorepo-flow.md reference.

Agent Routing Table

Task Execution (4-Step Cycle)

MANDATORY EXECUTION CYCLE: execute → branch on executor result → quality-fix → commit

Before the loop, register "Execute consumed task set", "Run post-implementation verification", "Clean up consumed task files", and "Report completion" once with TaskCreate; mark and advance the active phase with TaskUpdate.

For EACH task, YOU MUST:

1. EXECUTE: invoke Agent tool (subagent_type per routing table) → Record the current HEAD as diffBase, pass the task file path in the prompt, and receive the structured response
2. BRANCH ON EXECUTOR RESULT:
   • status: "escalation_needed" or "blocked" → STOP and escalate to user
   • requiresTestReview is true → Identify the changed integration/E2E test files in the current changes and invoke integration-test-reviewer with them as changedTestFiles, plus diffBase, taskFile, prompt-only claims, and mutationEvidence
     • approved → Proceed to step 3
     • blocked → STOP and escalate to user
     • needs_revision → Pass qualityIssues unchanged into the Review Resolution Gate; return to step 1 for rerouted corrections and derive convergence from correction re-review prior_feedback_reconciliation
   • status: completed → Proceed to step 3
3. QUALITY-FIX: Invoke the layer-appropriate quality-fixer with task_file, upstream mutationEvidence, and qualityCommand when available (caller first, otherwise current task)
   • stub_detected → Return to step 1 with incompleteImplementations[] details
   • blocked → STOP and escalate to user
   • approved → Proceed to step 4
4. COMMIT: Execute git commit after the layer-appropriate quality-fixer returns approved

CRITICAL: Parse every sub-agent response for status fields. Execute the matching branch in the 4-step cycle. Proceed to next task only after layer-appropriate quality-fixer returns approved.

Scope Boundary for Subagents

Append the following block to every subagent prompt invoked from this recipe:

Scope boundary for subagents:
Operate within the task scope and referenced files in the prompt.
Use loaded skills to execute that scope.
Escalate when the required fix or investigation falls outside that scope.

Verify task files exist per Pre-execution Checklist, then enter autonomous execution mode. When requirement changes are detected during execution, escalate to the user with the change summary before continuing.

Post-Implementation Verification (After All Tasks Complete)

Resolve all readable Design Docs from the Work Plan, or the Work Plan itself when none exist; missing input blocks verification.

Emit one code-verifier call per resolved document plus one security-reviewer call in one assistant message, then await all:

• code-verifier (subagent_type: "dev-workflows-fullstack:code-verifier") → each resolved doc_type, single document_path, and code_paths from git diff --name-only main...HEAD
• security-reviewer (subagent_type: "dev-workflows-fullstack:security-reviewer") → the typed governingDocuments list and implementationFiles

Apply subagents-orchestration-guide's Post-Implementation Verification pass/fail and fix/re-run rules with the layer-appropriate executor and quality-fixer. Present the unified report; proceed to Final Cleanup after all pass.

Final Cleanup

Before the completion report, delete the implementation task files this recipe consumed. Their work is committed; docs/plans/ is ephemeral working state and is not retained between recipe runs:

• Delete every file in the Consumed Task Set
• Preserve the work plan itself (docs/plans/{plan-name}.md) — the user decides whether to delete it after final review

If task files cannot be deleted (filesystem error), report the failure but do not block the completion report.

Completion Report Contract

Final report must include:

• Task materialization status
• Implemented task count, including backend/frontend counts
• Quality check result
• Commit count
• Cleanup result
• Declined actionable findings with ID, governing reason, and evidence, when any occurred
• Escalation or blocking summary, if any