packmind: Skill for Claude Code

Feature Sprint Skill

Execute a Packmind feature implementation plan via subagents with automatic parallel execution and sync-back. Companion to /feature-spec.

When to Use

Scenario	Use feature-sprint	Use plan + architect-executor
Fast execution of an already-specified feature	✅	❌
Parallel group execution	✅	partial
Multi-repo (OSS + proprietary) execution	✅	❌
Heavy TDD enforcement, per-task escalation	❌	✅
Specs in tmp/feature-specs/ (this flow)	✅	❌
Specs in .claude/specs/ and .claude/plans/	❌	✅

Rule of thumb: /feature-sprint is the executor for /feature-spec. For more careful, single-task-at-a-time execution with the global plan skill, use architect-executor instead.

Prerequisites

A completed feature spec at tmp/feature-specs/{slug}/:

• {slug}.md with status: COMPLETE in frontmatter
• context.md with YAML frontmatter (version 1.0+)
• implementation-plan.md with task checkboxes

If missing or DRAFT: tell the user to run /feature-spec {source} first.

Architecture: Orchestration + Subagents + Sync-Back

Core Principle: The main session orchestrates only. All implementation happens in Task subagents — that keeps the main context clean and lets agents do heavy reading without poisoning your conversation.

┌─────────────────────────────────────────────────────────────────┐
│  Persistent Layer (tmp/, git-ignored)                            │
│  ┌─────────────────────────────────────────────────────────────┐ │
│  │ implementation-plan.md                                      │ │
│  │ Source of truth: task checkboxes [ ] / [x]                  │ │
│  └─────────────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────┘
         │                              ▲
         │ Hydrate (session start)      │ Sync-back (per group)
         ▼                              │
┌─────────────────────────────────────────────────────────────────┐
│  Main Session (ORCHESTRATION ONLY)                               │
│  • Load spec & state         • Spawn subagents (one per group)  │
│  • Parse agent output        • Sync checkboxes                  │
│  • Run Nx quality gates      • Commit per group                 │
│  • NEVER implement tasks directly                                │
└─────────────────────────────────────────────────────────────────┘
                              │
                              │ Spawn (parallel or sequential)
                              ▼
┌─────────────────────────────────────────────────────────────────┐
│  Subagent Layer (IMPLEMENTATION)                                 │
│  ┌────────────────┐  ┌────────────────┐  ┌────────────────────┐ │
│  │ Group A agent  │  │ Group B agent  │  │ Group C agent      │ │
│  │ Backend tasks  │  │ Frontend tasks │  │ Tests / integration│ │
│  │ cwd: OSS repo  │  │ cwd: OSS repo  │  │ cwd: depends       │ │
│  └────────────────┘  └────────────────┘  └────────────────────┘ │
└─────────────────────────────────────────────────────────────────┘