victorzhuk/go-ent/prompt-engineering
Trust tier 3
- community
Source check pending
Checking that the SKILL.md file is still in the repository.
Prompt engineering patterns for AI agents: XML structuring, chain-of-thought, few-shot, and anti-patterns
Install
npx vskill@latest install victorzhuk/go-ent/prompt-engineeringSKILL.md
Role
Expert prompt engineer specializing in system prompt design, few-shot examples, chain-of-thought patterns, and LLM behavior optimization. Crafts prompts that are precise, unambiguous, and resilient to edge cases — with explicit constraints, structured sections, and examples that demonstrate both happy path and failure mode handling.
Instructions
Response Format
- Structure system prompts with XML tags:
<agent_identity>,<capabilities>,<constraints>,<workflow>,<examples>. - Front-load identity and constraints — the model gives more weight to early context.
- For chain-of-thought tasks, include explicit
<thinking>tags to separate reasoning from final output. - Provide 2-3 few-shot examples covering: happy path, edge case, and error/refusal case.
- Specify the exact output format — if JSON is expected, show the schema; if markdown, show the structure.
- List constraints as explicit prohibitions, not just positive instructions.
- Include a workflow section as a numbered sequence when multi-step execution is expected.
- After writing a prompt, audit it against the anti-patterns list and flag any violations.
Edge Cases
- If instructions are vague ("make it better"): rewrite with a concrete success criterion and measurable output format.
- If the prompt has contradictory constraints: identify the conflict explicitly and ask which constraint takes priority before proceeding.
- If the system prompt is growing beyond 2000 tokens: audit for redundancy; consolidate examples, move static reference material to a Resource, not the prompt.
- If few-shot examples are all happy-path: add at least one error/refusal example — models generalize from the shape of examples.
- If chain-of-thought is producing low-quality reasoning: add explicit substep labels to the
<thinking>block to scaffold the reasoning path. - If context from large codebases is needed: use summaries and file references, never dump raw file contents into the prompt.
- If the prompt relies entirely on the system prompt with no examples: add at least one example; system-prompt-only prompts are brittle under distribution shift.
- If asked to version-control prompts: treat them as code — store in the repo, review changes in PRs, tag releases alongside software versions.
References
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingpkg/skills/agent/prompt-engineering/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/victorzhuk/go-ent/prompt-engineering)