Skip to content
FSS Knowledge Base

Prompting

Section titled “Prompting”

Effective prompting is the foundation of AI productivity. Context engineering extends this into persistent, reusable knowledge.

Core Principles

Section titled “Core Principles”
  1. Be VERY clear on objectives — AI can only help with well-defined goals
  2. Start with an imperative verb: “Analyze…”, “Summarize…”, “Generate…”
  3. Strip filler (“please”, “thank you”) unless contextually important
  4. End important prompts with: “If you need additional information or clarification, ask.”
  5. Iterate on results — first draft is rarely final
  6. Review first draft of prompt and improve before sending

Frameworks

Section titled “Frameworks”

CLEAR Framework

Section titled “CLEAR Framework”
  • Context: background information and constraints
  • Length: desired output format and length
  • Examples: sample inputs/outputs
  • Audience: who will consume the output
  • Role: persona for the AI to adopt

Simple Structure

Section titled “Simple Structure”

Role -> Context -> Goal -> Output Rules

Perspective Prompting (Michael Simmons)

Section titled “Perspective Prompting (Michael Simmons)”

Five principles:

  1. Go beyond your own perspective
  2. AI is better at “perspectiving” than humans
  3. Power of multiple perspectives (expertise, stakeholder, temporal, outsider, failure-mode)
  4. Importance of different perspective types
  5. Synthesize tensions between perspectives — don’t average them

Key Techniques

Section titled “Key Techniques”

Meta-Prompting

Section titled “Meta-Prompting”

Multi-Prompting

Section titled “Multi-Prompting”
  • Send same prompt to 3 models (Claude, ChatGPT, Gemini)
  • Synthesize the AI responses to extract best insights
  • Cancels out individual model weaknesses

Verification Prompting

Section titled “Verification Prompting”
  • “Give the agent a way to verify its work”
  • Let the agent run tests, NOT edit tests (prevents gaming)
  • “Now critique this plan. What might go wrong?”
  • “Double-check everything. What might this have broken?”

Agentic Prompts

Section titled “Agentic Prompts”
Write a step-by-step plan to implement XXX. Then tell me what edge cases I should be aware of.
Now critique this plan. What might go wrong?
Break this into subtasks and spin up sub-agents to handle them in parallel.

Context Engineering

Section titled “Context Engineering”

CLAUDE.md as Context

Section titled “CLAUDE.md as Context”
  • Not documentation itself — a roadmap to documentation
  • Use @path/to/file to import context dynamically
  • Keep under 100 lines; reference external docs
  • See 01_Claude Code for memory system details

Shared Context Strategy

Section titled “Shared Context Strategy”
  • Research how to share context between Claude and Cursor
  • CLAUDE.md + .cursorrules in same project
  • Projects feature in Claude web: cache content, doesn’t count against limits

Session Management

Section titled “Session Management”
  • Keep sessions short and focused (~40 messages or one feature)
  • Use /compact every 30-40 messages to preserve context window
  • Name sessions for easy resume
  • Clear context between unrelated tasks

Prompt Library

Section titled “Prompt Library”
  • Store reusable prompts in .prompts/ folder or KB
  • Organize by task type: coding, research, strategy, writing
  • See Prompt Library subfolder for specific templates

Resources

Section titled “Resources”