AgentSkillsCN

executing-work-items

借助智能依赖分析与并行化能力，临时调用 AI 代理执行特定工作项。当您需要在正式冲刺之外单独实现一项或多项任务，或在测试单个功能的落地效果时，本技能都能助您灵活应对。

SKILL.md
--- frontmatter
name: executing-work-items
description: Executes specific work items ad-hoc using AI agents with intelligent dependency analysis and parallelization. Use when you need to implement one or more specific items outside of a formal sprint, or when testing implementation of individual features.
disable-model-invocation: true

Executing Work Items

Execute specific work items using AI agents with dependency resolution and parallel execution.

Usage

code
/executing-work-items <item-ids...> [options]

Arguments

Item IDs (required):

  • •One or more work item IDs (e.g., CCC-123 CCC-124 CCC-125)
  • •Or comma-separated: CCC-123,CCC-124,CCC-125

Options:

  • •--dry-run: Preview execution plan without running
  • •--force: Execute despite blockers or incomplete dependencies
  • •--skip-prep: Skip readiness validation

Examples

bash
# Execute single item
/executing-work-items CCC-123

# Execute multiple items
/executing-work-items CCC-123 CCC-124 CCC-125

# Preview plan only
/executing-work-items CCC-123 CCC-124 --dry-run

# Force execution (use with caution)
/executing-work-items CCC-123 --force

Workflow

Step 1: Gather Item Context

For each item:

python
item = pm_context.get_item(item_id)
item.children = pm_context.get_children(item_id)
item.parent = pm_context.get_item(item.parent_id) if item.parent_id else None

Step 2: Readiness Validation

Unless --skip-prep:

  • •Check all items have acceptance criteria
  • •Check for blocking dependencies
  • •Verify parent items exist (if specified)

Step 3: Dependency Analysis

python
# Check dependencies between items
for item in items:
    if item.blocked_by:
        blocker = pm_context.get_item(item.blocked_by)
        if blocker.status_type != "done":
            if blocker.id in item_ids:
                # Will execute in this batch - OK
                dependencies.append((blocker.id, item.id))
            else:
                # External dependency - warn
                warnings.append(f"{item.key} blocked by {blocker.key}")

Step 4: Code Analysis

Launch parallel analysis for each item:

  • •Identify files to modify
  • •Check for conflicts between items
  • •Plan implementation approach

Step 5: Plan Execution

Group items by:

  • •File overlap (same files → same agent)
  • •Dependencies (blocker runs first)
  • •Stack area (frontend, backend)

Assign to agents (max 4 parallel).

Step 6: Execute

Phase 1: Dependencies Execute blocking items first.

Phase 2: Parallel Run independent items in parallel.

Step 7: Report

markdown
## Execution Complete

### Items Executed: {count}/{total}

| Item | Status | Agent | Duration |
|------|--------|-------|----------|
{for item in items:}
| {item.key} | {status} | Agent-{n} | {duration} |

### Files Modified
{list of files}

### Next Steps
{any follow-up recommendations}

Use Cases

Ad-hoc Bug Fix

bash
/executing-work-items BUG-456

Feature Testing

bash
/executing-work-items FEAT-123 --dry-run
# Review plan, then:
/executing-work-items FEAT-123

Batch Task Execution

bash
/executing-work-items TASK-1 TASK-2 TASK-3 TASK-4

Invocation Control

This skill has disable-model-invocation: true because:

  • •Launches agents with side effects
  • •Modifies code and PM tool status
  • •User should explicitly control execution