Claude Code Plugins

Community-maintained marketplace

Feedback

task-management

@hhk7734/hhk7734
0
0

A comprehensive framework for high-performance engineering management and task execution. It guides users through reducing ambiguity, defining "Done," choosing the optimal leadership positioning, and managing risks via implication-based communication.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name task-management
description A comprehensive framework for high-performance engineering management and task execution. It guides users through reducing ambiguity, defining "Done," choosing the optimal leadership positioning, and managing risks via implication-based communication.

Task Management

This skill operationalizes "Seniority" and "Engineering Management" into a repeatable process. It enforces standards for reducing ambiguity, defining completion based on business value, and positioning oneself effectively between technical details and strategic direction.

1. Phase: Ambiguity Reduction (The Seniority Test)

True seniority is defined by the ability to take abstract/fuzzy requirements and turn them into concrete plans. Before execution, you must define the problem, not just the solution.

Diagnostic Questions:

  1. What is the underlying problem? (Separate the requested solution from the actual pain point).
  2. Who is the specific user? (Be specific; "the team" or "users" is insufficient).
  3. The "Why" Test: Can the engineer explain why this feature exists in the product vision? If the answer is "because PM asked," the context is broken.
  4. Risk Assessment: What happens if we are wrong?

Output Required:

  • A Problem Statement that replaces the original vague request.
  • Clarification Action Items (e.g., "Sync with stakeholders regarding naming conventions") [Conversation History].

2. Phase: Strategic Positioning (Command & Control)

Do not blindly "work hard." Determine your optimal positioning based on Situational Awareness (knowing what/why is happening) and Operational Clarity (team knows what to do).

Select Your Mode:

  • Crisis Mode (Low Awareness, Low Clarity): Learn & Stabilise. Prioritize coding/investigation to regain control immediately.
  • Ambiguity (High Awareness, Low Clarity): Lead by Example. Code alongside the team to set standards and build shared understanding.
  • Flying Blind (Low Awareness, High Clarity): Passive Coding. Trust the team's direction but do targeted contributions (e.g., bug fixes) to ramp up context.
  • Clarity (High Awareness, High Clarity): Strategic Direction. Step back from coding. Focus on long-term planning, risk mitigation, and "Wolf Time" (71/29 rule) allocation.

3. Phase: Definition of "Done" (Artifacts over Efforts)

"Done" is not a feeling or an effort; it is a social construct defined by the satisfaction of the stakeholder/company. Work is only complete when it produces readable results.

Standard for Engineering Completion:

Development is effectively "Done" only when the following artifacts exist:

  1. PR Merged: Code review passed and merged.
  2. CD Image: A deployable image generated via CI/CD.
  3. Versioned Helm Chart: A chart capable of running the image.
  4. End-to-End Validation: Proof that it works in the target environment (e.g., specific GPU targeting confirmed).

The "Done" Manifesto:

  • Do not report "Investigation" as a result. Report the Document produced.
  • Do not report "Refactoring" as a result. Report the Performance Metric improved or Tech Debt removed.
  • Declare Victory and Leave: When the criteria are met, explicitly state "This task is complete" and move to the next challenge. Do not get trapped in infinite gardening.

4. Phase: Execution & Communication

Communication must bridge the gap between technical facts and business decisions.

Implication-Based Communication:

  • BAD (Fact-only): "OOM occurred." / "Cache hit rate changed."
  • GOOD (Implication): "OOM occurred, which implies we cannot support the target batch size. Recommendation: Decrease batch size or increase GPU memory request."

Risk Management:

If the "Expected Result" and the "Schedule" are misaligned:

  1. Acknowledge the gap immediately.
  2. Identify the cause.
  3. Propose a mitigation plan (e.g., "Draft by Jan 7, Final by Jan 14").

Example Usage

Input Task: "Action item: Please actually write the templates and validate that they work end-to-end."

Applied Framework:

  1. Ambiguity: Clarified "Write templates" -> "Create Odin presets for specific GPU models." Synced on naming conventions.
  2. Positioning: Ambiguity Mode. The manager/senior engineer will write the initial templates (Lead by Example) to set the standard for the Hanoi team.
  3. Definition of Done: Artifacts = Preset File + Doc + Validation Log + Versioned Chart.
  4. Closing: "Task Complete. Hanoi team can now target GPUs using label x. Documentation is at link."