/sc-work

sc-work: Universal Implementation Command

Purpose: I need to build/fix/modify something - handles all implementation tasks from creating new features to fixing bugs to refactoring code

Agent Orchestration and Deployment Strategy

Efficiency First: Handle tasks directly when possible. Use agents only when genuinely needed for:

• Multi-step coordination requiring handoffs
• Specialized domain expertise beyond general capability
• Parallel work streams with interdependencies
• Complex analysis requiring multiple perspectives
• Operations that produce verbose intermediate output

Direct Agent Rules (ALWAYS delegate these):

• Documentation lookups → Use sc-repo-documentation-expert
• Test execution → Use sc-test-runner
• Web searches → Use sc-web-search-researcher
• GitHub searches → Use sc-github-researcher
• Deep codebase analysis → Use sc-code-explorer
• Architecture design → Use sc-code-architect
• Code quality review → Use sc-code-reviewer

Available Agents:

• sc-code-explorer - Deeply analyzes existing codebase features by tracing execution paths and mapping architecture
• sc-code-architect - Designs feature architectures by analyzing existing patterns and providing implementation blueprints
• sc-code-reviewer - Reviews code for bugs, security vulnerabilities, and adherence to project conventions
• sc-repo-documentation-expert - Finds documentation from Context7, local repos, and GitHub repositories
• sc-test-runner - Executes tests and analyzes failures
• sc-web-search-researcher - Searches web for current information
• sc-github-researcher - Searches GitHub for repos, code, issues, and PRs via gh CLI

Context Preservation:

• Keep only: user request, actionable recommendations, code changes, summary, next steps
• Discard: intermediate outputs, full docs, verbose logs, exploratory reads

Processing Pipeline: Parse → Classify → Validate → Route → Execute → Synthesize

Intent Recognition and Semantic Transformation

This command interprets natural language requests that express the intent: "I need to build/fix/modify something" - implementation work including creating, modifying, fixing, refactoring, optimizing, or enhancing code.

Command Execution:

Empty $ARGUMENTS: Display usage suggestions → stop Has content: Parse intent → apply strategy → route execution

Intent Processing: Extract intent → Apply strategy matrix → Validate → Execute

Strategy Matrix:

Condition	Direct Handling	Agent Required
Task Type	Simple, single-step	See "Direct Agent Rules" above
Domain	Single, familiar	Multi-tech, unknown
Context	Available locally	External research needed
Output	Concise, focused	Verbose, needs filtering

Transforms: $ARGUMENTS into structured execution:

• Intent: [implementation-goal-and-scope]
• Approach: [direct-implementation OR research-then-implement]
• Agents: [none OR minimal-viable-set]

Output Template

## Response
[Direct answer or action taken - 1-3 sentences addressing the core request]

## Details
[Main content based on command type:
- Plan: Strategy breakdown with phases
- Work: Code changes and implementation steps
- Explore: Research findings and analysis
- Review: Issues found and quality assessment]

## Next Actions
[What to do next:
- Plan: Implementation steps to begin
- Work: Testing and validation needed
- Explore: Areas for deeper investigation
- Review: Fixes and improvements to make]

## Notes
[Optional - context, warnings, alternatives, or additional considerations]

---

User Request: $ARGUMENTS