| name | prompt-engineering |
| description | Analyze and optimize prompts for AI coding agents. Decompose tasks into components (skills/commands/agents), detect missing context, and produce ready-to-paste improved prompts. Advisory only — outputs prompts, not code. Triggers: "optimize prompt", "improve my prompt", "rewrite this prompt", "how to write a prompt for", "help me prompt"
|
Prompt & Context Engineering
Analyze a draft prompt, identify gaps, map it to available agent components,
and output an optimized prompt the user can paste and run.
2026 note: The field has evolved from "prompt engineering" to "context engineering" — managing the full context window (system prompts, retrieved data, tool schemas, conversation history) rather than just optimizing the wording of a single prompt.
When to Use
- User asks to optimize, improve, or rewrite a prompt
- User asks "what's the best way to ask the agent to..."
- User pastes a draft prompt and asks for feedback
When to Skip
- User wants the task executed directly ("just do it")
- User asks to optimize code or performance (that's a refactoring task)
How It Works
Advisory only — output analysis and an optimized prompt. Do not execute the task.
Phase 1: Project Detection
Before analyzing the prompt, detect context from the working directory:
- Read
CLAUDE.md / AGENTS.md for project conventions
- Detect tech stack from config files:
package.json → Node / TypeScript / React / Next.jsgo.mod → Gopyproject.toml / requirements.txt → PythonCargo.toml → Rustbuild.gradle / pom.xml → Java / Kotlin
- Note the detected stack for component matching and missing-context analysis
If no project files found, flag "tech stack unknown" and proceed.
Phase 2: Intent Classification
Classify the user's task:
| Category | Signal Words | Example |
|---|
| New Feature | build, create, add, implement | "Build a login page" |
| Bug Fix | fix, broken, not working, error | "Fix the auth flow" |
| Refactor | refactor, clean up, restructure | "Refactor the API layer" |
| Research | how to, what is, explore | "How to add SSO" |
| Testing | test, coverage, verify | "Add tests for the cart" |
| Review | review, audit, check | "Review my PR" |
| Documentation | document, update docs | "Update the API docs" |
| Infrastructure | deploy, CI, docker, database | "Set up CI/CD pipeline" |
Phase 3: Scope Assessment
| Scope | Heuristic | Approach |
|---|
| Trivial | Single file, < 50 lines | Direct execution |
| Low | Single component or module | Single skill or tool |
| Medium | Multiple components, same domain | Chained steps + verification |
| High | Cross-domain, 5+ files | Plan first, then phased execution |
| Epic | Multi-session, architectural shift | Multi-session plan with checkpoints |
Phase 4: Component Mapping
Map intent + scope to available agent components:
| Intent | Skills | Agents |
|---|
| New Feature | dev, dev-frontend/backend, dev-scaffolding | planner, code-reviewer |
| Bug Fix | dev-testing, dev-debugging | tdd-guide |
| Refactor | dev-code-reviewer | code-reviewer |
| Research | search, rag | explore agent |
| Testing | dev-testing | test runner |
| Review | dev-code-reviewer | code-reviewer |
| Documentation | documentation | doc writer |
| Infrastructure | dev-backend | architect |
For project-specific skills, check .agents/skills/ to find applicable ones.
Phase 5: Missing Context Detection
Scan the prompt for missing information:
If 3+ items are missing, ask the user up to 3 clarification questions before
generating the optimized prompt.
Phase 6: Generate Optimized Prompt
Produce two versions:
Full version (inside a fenced code block):
- Clear task description with context
- Tech stack (detected or specified)
- Acceptance criteria
- Verification steps
- Scope boundaries (what to exclude)
- Relevant skills/agents to invoke
Quick version (one-liner patterns):
| Intent | Pattern |
|---|
| New Feature | Plan [feature]. Implement with tests. Review. Verify. |
| Bug Fix | Write failing test for [bug]. Fix to green. Verify. |
| Refactor | Refactor [scope]. Review. Verify no regressions. |
| Research | Search for [topic]. Summarize findings with citations. |
| Testing | Add tests for [module]. Target [coverage]% coverage. |
Prompt & Context Quality Principles
- Specificity over vagueness: name files, modules, endpoints
- Acceptance criteria: define "done" before starting
- Scope boundaries: state what to exclude to prevent drift
- Existing patterns: reference existing code the agent should follow
- Verification step: end with a concrete check (test, build, curl)
- Component awareness: invoke skills/agents by name when applicable
- Context assembly: include relevant code, docs, and data — not just the question
- Self-reflection: ask the model to verify its own reasoning (CoT) before final output
- Treat prompts like code: version control prompts, run evals, iterate based on metrics
- Tool-augmented prompting: when a prompt involves factual lookup, direct the model to use search/MCP tools rather than relying on training data
Output Format
### Prompt Diagnosis
**Strengths**: (what the original does well)
**Issues**: (table: issue | impact | fix)
**Needs Clarification**: (numbered questions)
### Recommended Components
(table: type | component | purpose)
### Optimized Prompt — Full
(fenced code block, ready to paste)
### Optimized Prompt — Quick
(one-liner)
### Enhancement Rationale
(table: what was added | why)
Constraints
- Advisory only — produce prompts, not implementations
- Respond in the same language as the user's input
- When referencing skills, verify they exist in
.agents/skills/ first
- Keep optimized prompts self-contained and copy-pasteable
