| name | quick-build |
| description | Lightweight design-and-implement skill for small-to-medium tasks. Collapses brainstorm + plan + build into one session. |
| user-invocable | true |
| disable-model-invocation | true |
| allowed-tools | Read, Edit, Write, Grep, Glob, Bash, AskUserQuestion, WebSearch, mcp__context7__resolve-library-id, mcp__context7__query-docs, mcp__sequential-thinking__sequentialthinking, TaskCreate, TaskUpdate, TaskList |
Quick Plan
Design and implement small-to-medium tasks in a single session — no intermediate files, no skill handoffs.
Announce at start: "I'm using the quick-build skill to design and implement this task."
Core Principles
Design philosophy:
- YAGNI — Push back on speculative features. If there's no concrete need today, leave it out.
- KISS — Bias toward the simplest solution that works. Challenge unnecessary complexity.
- DRY — Reuse existing patterns and modules in the codebase before proposing new ones.
- Defensive Engineering — Assume failures are possible. Right-size risk analysis to the feature’s complexity. Surface the most likely bottlenecks, silent failures, and edge cases for this domain.
Conversation style:
- Always use AskUserQuestion — Use the AskUserQuestion tool with
Other options for all questions to the user.
- One question at a time — Never ask multiple questions in a single message.
- Multiple choice when possible — Easier to answer than open-ended when possible
- Explore before committing — Always present 2–3 approaches before settling on one.
- Incremental validation — Get user approval before moving between phases.
- Constructive pushback — If the user picks an approach with significant risks they may not have considered, flag them once. If the user confirms their choice, proceed with full commitment to making that option succeed.
Workflow
Phase 1: Context & Discovery
Step A — Read the codebase first:
Tell the user "Let me explore the relevant parts of the codebase first." then do the following silently (don't narrate each file read):
- Quick scan — Understand the relevant area: project structure, key modules, existing patterns.
- Locate the target — Identify where the change belongs. If the location or integration flow is unclear, this becomes your first question to the user.
Step B — Scope the problem:
If the task is too broad or complex for a single session, suggest escalating to /brainstorm for the full design flow. Indicators of complexity:
- Requires significant architectural decisions with long-term implications
- Has unclear requirements that need extended discovery
Step C — Ask questions to fill the gaps:
Ask questions one at a time to deeply understand the purpose, constraints, and success criteria.
Keep going until you can clearly articulate: what we're building, why it matters, and where it fits in the existing system.
Phase 2: Propose Options & Trade-offs
Present 2–3 viable approaches. Lead with your recommendation and why.
For each option, use this format:
## Option X: [Name] (Recommended — if applicable)
**Approach**: [Brief description]
**When to choose**: [One-liner on what makes this the right pick]
**Pros**:
- [Advantage 1]
- [Advantage 2]
**Cons**:
- [Disadvantage 1]
- [Disadvantage 2]
**Effort**: [Low/Medium/High]
**Risk**: [Low/Medium/High]
**Fits existing patterns**: [Yes/Partially/No]
After listing all options, include a quick comparison table:
| Criteria | Option A [Name] | Option B [Name] | Option C [Name] |
| ------------- | --------------- | --------------- | --------------- |
| Effort | Low | Medium | High |
| Risk | Medium | Low | Low |
| Fits patterns | Yes | Partially | No |
Wait for the user to pick one before moving to Phase 3.
Phase 3: Inline Plan & Execute
Step A — Present the plan:
Present a concise implementation plan directly in the conversation. No file is saved. Format:
## Implementation Plan
### 1. [Action description]
[Brief description of what changes and why]
### 2. [Action description]
[Brief description of what changes and why]
### 3. Verify
[How to confirm it works — tests, commands, etc.]
Wait for user approval before writing any code.
Step B — Implement:
Execute the plan directly:
- Make the code changes
- Run verifications (tests, lint, build — whatever is relevant)
- Report results
If you hit a blocker during implementation, stop and ask rather than guessing.
Research
- Use WebSearch for general research: evaluating approaches, exploring best practices, or comparing established patterns.
- Use Context7 for library-specific documentation: version-specific APIs, configuration options, or implementation examples.