reference

Prompt Patterns Library — Reusable Prompts for Collaboration Tasks

A concise, practical library of tested prompt patterns for meeting notes, decision synthesis, email drafting, agenda generation, and retrospectives. Each pattern includes purpose, a ready-to-use example, useful variants, expected output format, and a verification checklist plus governance metadata and guardrails for safe, consistent AI-assisted collaboration.

Prompt Patterns Library — Reusable Prompts for Collaboration Tasks

Purpose: Give teams safe, high-value starter prompts they can adapt and store so AI assists consistently across meetings, decisions, email, and facilitation work. Each pattern below shows why to use it, a copy-ready example, practical variants, the expected output shape, and a short verification checklist for quality and safety.

How to use these patterns

  • Copy an example and substitute the bracketed variables (e.g., [PROJECT], [AUDIENCE]).
  • Edit tone, length, or format requirements to match your organization's style guide.
  • Run the output through the verification checklist before sharing or acting on it.
  • Store adapted prompts in your team prompt library with metadata (intent, owner, tags, last-reviewed, human-review-required).

Standard metadata to store with every prompt

  • Intent: Short description of the outcome the prompt should produce.
  • Owner: Person responsible for maintaining the prompt.
  • Tags: e.g., meeting, email, decision, retrospective.
  • Last reviewed and Review cadence.
  • Human review required? (yes/no)
  • Bias/risks notes: known issues and guardrails.

Patterns

1) Meeting Summary — Concise Notes + Action Items

Purpose: Create clear meeting notes that are scannable and action-oriented for distribution.

Example prompt:

"You are a concise meeting summarizer. Given the raw meeting transcript below, produce: (a) a 3–5 sentence executive summary, (b) a bulleted list of decisions made, (c) a prioritized action list with owners and due dates (if mentioned), and (d) any open questions. Keep language neutral and suitable for distribution to stakeholders. Transcript: [PASTE TRANSCRIPT]"

Variants: include a "Technical Appendix" section, produce a one-sentence TL;DR, or format as email-ready copy.

Expected output format: HTML or Markdown with clearly labeled sections (Summary, Decisions, Actions, Open questions).

Verification checklist:

  • Are all named action owners correct and spelled consistently?
  • Are due dates present or marked "TBD" when missing?
  • Does the summary avoid adding new claims not in the transcript?

2) Decision Synthesis — From Options to Recommendation

Purpose: Synthesize discussion and criteria into a compact decision brief and recommended next step with rationale and risks.

Example prompt:

"You are a decision analyst. Given the decision context, options, evaluation criteria, and relevant facts below, produce: (1) a 1-paragraph recommendation with rationale, (2) a short pros/cons table for each option, (3) key assumptions, and (4) one contingency plan if the recommendation fails. Context: [CONTEXT]. Options: [LIST]. Criteria: [CRITERIA]. Facts: [FACTS]."

Variants: produce a risk matrix, or generate a 90-second spoken summary for leadership briefings.

Expected output format: Structured brief with sections labeled Recommendation, Pros/Cons, Assumptions, Contingencies.

Verification checklist:

  • Does the recommendation reference the stated criteria?
  • Are risks and assumptions explicit and plausible?
  • Is any missing or uncertain data flagged for human follow-up?

3) Email Draft + Tone Adjustments

Purpose: Draft professional emails and adapt tone (friendly, formal, brief, persuasive) while keeping brand voice consistent.

Example prompt:

"Draft an email to [RECIPIENT ROLE] about [TOPIC]. Include a polite opener, one-paragraph explanation, a clear call to action, and a closing. Tone: [TONE]. Keep it under 200 words. Company voice: [BRAND VOICE NOTES]."

Variants: create follow-ups, polite declines, or escalation templates.

Expected output format: Ready-to-send email body with subject line suggestion.

Verification checklist:

  • Is the requested tone consistent?
  • Does the CTA clearly state next steps and timing?
  • Are any sensitive details omitted or redacted as required?

4) Agenda Generation — Time-Boxed, Outcome-Focused

Purpose: Turn a meeting purpose, participant list, and available time into a time-boxed agenda with facilitator cues.

Example prompt:

"Create a 45-minute agenda for a meeting with these participants: [LIST]. Purpose: [PURPOSE]. Desired outcomes: [OUTCOMES]. Include time boxes, lead person for each item, and one 5-minute decision-check at the end."

Variants: icebreakers for new teams, pre-read recommendations, or asynchronous alternatives.

Expected output format: Bulleted agenda with timestamps, owners, and desired output per item.

Verification checklist:

  • Are times realistic given the topic scope?
  • Does every agenda item have a clear desired output?
  • Is there an explicit decision or next step at the end?

5) Retrospective Facilitation Prompts

Purpose: Help facilitators structure safe, productive retrospectives that surface actionable improvements.

Example prompt:

"You are a retrospective facilitator. Given these sprint highlights and metrics: [HIGHLIGHTS], produce: (a) three short activities to surface what went well, what to improve, and experiments to try; (b) example facilitation scripts for each activity; and (c) a prioritized improvement backlog with owners."

Variants: remote-friendly formats, anonymous input collection prompts, or lightweight sprint blameless postmortems.

Expected output format: Facilitator script, activities list, prioritized backlog.

Verification checklist:

  • Do prompts avoid blaming language and encourage psychological safety?
  • Are proposed experiments measurable and time-boxed?

Governance & Guardrails

Store every adapted prompt in the team library with the standard metadata above. Mark prompts that require mandatory human review (e.g., legal, regulatory, clinical) and route outputs to the appropriate reviewer before distribution.

Bias and safety guardrails:

  • Explicitly request sources and evidence when outputs make factual claims.
  • Ask the model to flag uncertainty ("If you are unsure, say 'uncertain' and list needed facts").
  • Use conservative phrasing for sensitive topics and require human sign-off.

Verification & Versioning

Keep a simple version history and change log for any prompt used for operational tasks. Include who changed it, why, and a short test example showing before/after output quality. Review prompts at a regular cadence (quarterly or aligned to risk level).

Quick starter checklist for adopting this library

  1. Pick 3 high-value patterns your team will rely on.
  2. Customize examples with your brand voice and required fields.
  3. Store adapted prompts with metadata and assign an owner.
  4. Define who must perform human review and when.
  5. Log at least one verification run and approve before broad use.

Note: This library is a living resource. Treat patterns as starting points—measure output quality, capture common failures, and iterate your stored prompts over time.

Discuss Ask the Librarian

Discussion

Comments and conversation will live here.