PILOT — Private preview. Progress is saved for this browser session only.
HaiPhai.AI Fluency for Biotech

Building a Shared Team Prompt Library

Lesson 4~16 min1-question check

Module 11 · Lesson 04

Building a Shared Team Prompt Library

Reading time: 16 minutes Track: Claude Fluency for Teams · Required for all learners


The compounding value problem

When a team member writes a great prompt for a recurring task, that knowledge typically dies in one of three places:

  1. Buried in their conversation history
  2. Living in their personal notes
  3. Transferred only if someone happens to ask

This means every team member reinvents prompts for the same tasks, quality varies wildly between people, and good work doesn't compound. A prompt library solves this.

What belongs in a prompt library

Not every prompt is worth capturing. The criteria for library-worthy prompts:

  • Recurring: Used more than 2-3 times per month
  • Non-trivial: Took meaningful effort to get right
  • Transferable: Works for others doing the same task, not just the original author
  • Verified: Produced actually good output, not just okay output

Common categories worth capturing:

CategoryExamples
Code tasksPR review template, debugging starter, test generation
Document tasksEmail templates, report structure, executive summary
AnalysisCompetitive analysis framework, decision matrix, risk analysis
CommunicationMeeting follow-up, stakeholder update, incident post-mortem
ResearchLiterature review, technology comparison, requirement extraction

Anatomy of a library entry

A useful prompt library entry contains more than just the prompt. Each entry should have:

Prompt name: Short, searchable label Use case: One sentence on when to use this The prompt: Full text, with [PLACEHOLDERS] for variable parts Example output: A real output this produced (helps others evaluate fit) Notes: What makes this prompt work, common pitfalls, suggested variations Owner/date: Who wrote it, when — so you know who to ask questions

Example entry:


Name: Technical Post-Mortem Draft

Use case: After any production incident, generate a structured post-mortem outline

Prompt:

You are a senior engineer writing a blameless post-mortem for [INCIDENT DESCRIPTION].

Timeline of events: [PASTE TIMELINE]
Services affected: [LIST]
Customer impact: [DESCRIBE]

Write a structured post-mortem covering:
1. Summary (2-3 sentences)
2. Timeline (format as table: time | event | who noticed)
3. Root cause analysis (what happened, not who caused it)
4. Contributing factors
5. Impact summary
6. Remediation actions (format as: action | owner | due date)
7. Preventive measures for the future

Tone: factual, blameless, focused on systems not individuals.

Notes: Works best when you provide a detailed timeline. The more specific the incident description, the less generic the output.


Where to keep the library

The format matters less than it being used. Common options:

  • Notion or Confluence page: Easy to update, comment on, share links. Good for teams already using these tools.
  • GitHub/GitLab repo: Prompts as markdown files, version-controlled, PRs for additions. Good for engineering teams.
  • Google Docs folder: Simple, accessible. Works for any team.
  • CLAUDE.md sections: For project-specific prompts, a "Common Prompts" section in CLAUDE.md puts them directly in Claude Code sessions.

Maintaining the library

A prompt library that isn't maintained becomes useless. Two practices keep it alive:

  1. Monthly review: 30-minute team session to add new prompts, mark outdated ones, update prompts as the team's tools or needs change.
  2. Usage signals: Note in each entry the last time it was used. Prompts unused for 6 months should be archived or deleted.

Getting started: the 3-prompt sprint

Don't try to build a comprehensive library from day one. Instead, run a 3-prompt sprint with your team:

  1. Each person identifies one recurring task where they've developed a good prompt
  2. Everyone writes up that prompt using the template above
  3. Share and debrief: what works, what's missing

Three solid, well-documented prompts that people actually use beat fifty prompts nobody looks at.

Knowledge check

1 question · select an answer to see if you got it
1.A team member writes a great prompt for generating incident post-mortems. What should happen next?
Ready to apply this?
Practice with AI →

Bring a real challenge from your work — the AI will help you apply what you just learned.