AgentSkillsCN

interactive-plan

通过决策性问题、任务列表、整洁的代码结构,以及测试策略,实现交互式生产计划。

SKILL.md
--- frontmatter
name: interactive-plan
description: Interactive production planning with decision questions, task list, clean code, and test strategy.

Interactive Planning Skill

Goal

Create a production implementation plan using interactive decision steps.

This skill:

  • Detects missing decisions
  • Asks structured questions
  • Accepts multiple choice or free-text answers
  • Builds a final implementation plan
  • Generates a task checklist
  • Enforces quality and testing expectations

Execution Mode

Default:

  • Deterministic
  • Risk-aware
  • Minimal hallucination
  • Repo-aware if context exists
  • Tool-aware if scripts/tests/CI exist

Workflow


Phase 1 — Context Scan

If repository context exists, quickly inspect:

  • README / docs
  • Architecture files
  • Test setup
  • CI / CD commands

Detect if possible:

  • Language / framework
  • Test framework
  • Deployment model
  • Risk areas

If tests exist:

  • Prefer extending existing tests over inventing new structure.

Phase 2 — Decision Question Mode

If planning uncertainty exists, ask up to 3 questions.

Question Format (MANDATORY)

code

## Decision Needed

### Q<n> — <Decision Topic>

Why this matters: <Short explanation>

Choose one:
A) Option A
B) Option B
C) Option C
D) Other: _______


Answer Handling

User may respond with:

  • Single letter (A / B / C)
  • Free text
  • Mixed answers

If user does not answer:

  • Choose safest production default
  • Record as assumption

Phase 3 — Plan Generation

After decisions are resolved (or safely assumed), generate final plan.


Quality Requirements


Clean Code Principles

Plan must encourage:

  • Small functions
  • Clear naming
  • Single responsibility
  • Testability by design
  • No hidden side effects
  • Explicit error handling

Testing Requirements

Must consider:

Unit Tests

Business logic validation

Integration Tests

System boundary behavior

Failure Tests

Timeouts
Retries
Invalid input
Dependency failures

Regression Tests

If modifying existing behavior


Observability (if backend or infrastructure)

Consider:

  • Structured logging
  • Metrics
  • Alert triggers
  • Health checks

Output Template

code

# Implementation Plan

## Approach

<1–3 sentences describing approach>

---

## Scope

### In Scope

*

### Out of Scope

*

---

## Decisions

| Decision   | Choice            | Reason |
| ---------- | ----------------- | ------ |
| <Decision> | <Selected Option> | <Why>  |

---

## Action Items

[ ] Discovery / validation
[ ] Implementation step
[ ] Tests added
[ ] Risk validation
[ ] Rollout strategy

---

## Task Checklist (Detailed)

| Task                     | Risk                | Confidence          |
| ------------------------ | ------------------- | ------------------- |
| <Verb-based atomic task> | Low / Medium / High | Low / Medium / High |

---

## Test Strategy

Unit:
Integration:
Edge cases:
Failure scenarios:

---

## Risks

Risk:
Mitigation:

---

## Assumptions

*

---

## Open Questions (if any)

*


Planning Rules

Prefer:

  • Vertical slices
  • Safe incremental rollout
  • Feature flags for risky changes
  • Backward compatibility
  • Observable deployments

Avoid:

  • Big bang rewrites
  • Hidden coupling
  • Untestable designs
  • Large speculative refactors

Failure Rules

If context missing → Ask
If decision unclear → Choose safest default and record assumption
If risk high → Highlight explicitly
If estimate uncertain → Mark clearly


Non Goals

Do NOT:

  • Write full production code
  • Generate full specifications unless explicitly requested
  • Redesign full architecture unless explicitly requested