Imported from akillness/jeo-skills (.agent-skills/prompts-chat/SKILL.md). Install upstream with npx skills add akillness/jeo-skills --skill prompts-chat. Copyright stays with the author (MIT).

prompts-chat — Curated AI Prompt Discovery & Application

Discover and apply high-quality prompts from the prompts.chat collection to enhance AI interactions, workflows, and prompt engineering.

Reference: prompts.chat — a community-driven repository of useful prompts for various AI tasks and workflows.

When to use this skill

• You need a well-crafted prompt template for a specific task or domain
• You're optimizing an AI workflow and want proven prompt patterns
• You're building a prompt library or managing prompt versions
• You want to discover prompts for roles, tasks, or use cases you're unfamiliar with
• You're troubleshooting response quality by refining the prompt structure

When not to use this skill

• The main job is building a full LLM application framework → use pydantic-ai, crewai-multi-agent, or langgraph-human-in-the-loop
• The main job is prompt optimization via RLHF or fine-tuning → use dspy, openrlhf-training, or moe-training
• The main job is building an agent system → use deep-agents-core, crewai-multi-agent, or workflow skills like harness
• The main job is RAG pipeline construction → use llamaindex or supabase-agent-skills

Instructions

Step 1: Classify your prompt need

Identify the category and use case:

prompt_need:
  category: writing | coding | analysis | brainstorming | role-play | instruction-tuning | other
  use_case: single-task | template-library | workflow-pipeline | prompt-versioning
  domain: software | marketing | education | science | creative | business | other
  current_pain: quality | consistency | structure | discovery | version-control

Step 2: Fetch and explore prompts.chat data

Access the prompts.chat repository to find relevant prompts:

# Clone or fetch the latest prompts collection
git clone https://github.com/f/prompts.chat.git /tmp/prompts-chat 2>/dev/null || \
  curl -s https://api.github.com/repos/f/prompts.chat/contents/ | jq '.[] | select(.type=="dir") | .name'

# List available categories
ls /tmp/prompts-chat 2>/dev/null || echo "Use online repository"

Step 3: Evaluate and select prompts

For each candidate prompt, assess:

1. Relevance: Does it match your task or domain?
2. Quality: Is the prompt structure clear and well-organized?
3. Adoptability: Can you integrate it into your workflow without modification?
4. Reusability: Does it work as a template for variations?

Step 4: Adapt prompts to your context

Customize selected prompts:

• Replace placeholders with your specific context
• Adjust tone, length, or output format to fit your workflow
• Test on representative inputs before full deployment
• Document variations and version changes

Step 5: Store and manage prompt versions

Create a local prompt library for your project:

# Example structure
prompts/
  ├── approved/
  │   ├── code-review.md
  │   ├── technical-writing.md
  │   └── summarization.md
  ├── drafts/
  └── VERSIONS.md  # Track changes

Document:

• Source (prompts.chat or custom)
• Version date
• Intended use case
• Known limitations or tips

Step 6: Integrate into workflows

Embed prompts into agent workflows:

# Reference prompt files in scripts
PROMPT=$(cat prompts/approved/technical-writing.md)
curl -s https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -d @- <<EOF
{
  "model": "gpt-4",
  "messages": [{"role": "system", "content": "$PROMPT"}]
}
EOF

Step 7: Monitor and iterate

Track prompt performance:

• Note which prompts yield the best outputs
• Collect feedback from users or automated quality metrics
• Refine based on performance data
• Share improvements back to the community (optional)

Examples

Example 1: Finding a code-review prompt

Goal: Improve code review quality using a template prompt

Search prompts.chat for "code review" or "code analysis" category
→ Find the "Code Review" prompt
→ Adapt it with your repo's standards and style guide
→ Store in `prompts/approved/code-review.md`
→ Use in code review workflows

Example 2: Building a prompt library for content creation

Goal: Create consistent content across multiple topics

Collect prompts for:
  - SEO-optimized blog post writing
  - Social media caption generation
  - Newsletter content curation
  - Video script structuring
→ Store each in prompts/approved/
→ Version and document in VERSIONS.md
→ Integrate into content pipeline

Example 3: Prompt versioning and A/B testing

Goal: Compare prompt effectiveness

Create versions:
  - prompts/approved/summarization-v1.md (original)
  - prompts/approved/summarization-v2.md (refined)
→ Test both on sample inputs
→ Measure quality metrics (brevity, accuracy, completeness)
→ Keep best version, document learnings

Best practices

1. Start with curated sources — prompts.chat is battle-tested; use it as a foundation
2. Document your source — track which prompts come from where and why
3. Version your prompts — treat prompts like code; version and track changes
4. Test before deploying — validate new or modified prompts on representative inputs
5. Share learnings — contribute back to communities like prompts.chat when you improve a prompt
6. Separate by use case — organize prompts by domain, role, or task for discoverability
7. Keep feedback loops — collect data on which prompts work best and iterate
8. Avoid prompt bloat — retire or consolidate underused prompts regularly

Integration with other skills

• dspy — Use prompts.chat prompts as seed optimizers for DSPy pipelines
• pydantic-ai — Embed curated prompts into Pydantic AI agent systems
• crewai-multi-agent — Structure multi-agent teams with role-specific prompts
• llm-wiki — Store and version prompts in your durable knowledge base
• technical-writing — Use prompts to help structure documentation workflows

References

• prompts.chat GitHub
• OpenAI Prompt Engineering Guide
• Anthropic Prompt Writing Guide
• Prompt Engineering Guide (community)