Peer & Team Professional Recommendation
Writes a LinkedIn recommendation for someone you've worked with, grounded in the real evidence of their work pulled from your connected tools -- not generic praise. Gives you three length variants plus a checklist to verify every claim before posting.
---
name: peer-team-recommendation
description: "Writes a personalized LinkedIn recommendation for someone the user has worked with -- a report, peer, manager, or cross-functional partner -- grounded in real evidence of that person's work pulled from connected tools (Gmail, Calendar, Drive, Slack). Trigger when the user asks to 'write a LinkedIn recommendation', 'recommend someone on my team', 'write a recommendation for [name]', 'endorse a colleague', or 'write a reference for [name]'. Gathers the few facts it needs interactively, mines the evidence, and returns three length variants plus an evidence checklist so the user can verify before posting."
---
# Peer & Team Professional Recommendation
Help the user write a LinkedIn recommendation for someone they've worked with. The recommendation has to sound like the user wrote it and be grounded in real, specific evidence of the person's work -- not generic praise with a name swapped in. The whole value is specificity backed by things that actually happened.
The job has two halves: gather the evidence from the user's connected tools, then write drafts in the user's voice that a reader will find credible because they're concrete.
## Step 1 -- Gather inputs
Collect what you can from the user's opening message first, then ask only for what's missing. Don't re-ask for anything they already gave you.
Required:
1. **Who** -- the person's name.
2. **Relationship** -- the user's relationship to them: their manager, their direct report, a peer, or a cross-functional partner. This changes which evidence matters and how the recommendation should sound.
Helpful, ask if not obvious:
3. **Their role / title** -- so the recommendation frames them correctly.
4. **People leader?** -- do they manage a team or have direct reports? If yes, look for leadership evidence (coaching, hiring, developing people, scaling impact through others). If no, skip that whole search.
5. **Anything to emphasize or avoid** -- a specific project the user wants featured, a strength they want to lead with, or something to steer clear of. Optional.
Sensible defaults, mention them rather than asking:
- **Review period** -- last 12 months unless the user says otherwise.
- **Region** -- only matters if it's needed to disambiguate the person or frame their scope. Ask only if relevant.
If the user only gives you a name and "write a recommendation," ask for the relationship and whether the person is a people leader before searching. Those two answers shape everything downstream.
## Step 2 -- Mine the evidence
Use the connected tools available (Gmail, Google Calendar, Google Drive, Slack, or whatever is connected) to find this person's most meaningful work over the review period. On ChatGPT this is Company Knowledge; on Claude it's the connected Gmail / Calendar / Drive / Slack tools. Use whatever is wired up; if nothing is connected, say so and ask the user to paste in what they have.
Look for:
- Major projects, launches, outcomes, and customer impact they drove or contributed to.
- Leadership moments and examples of collaboration across teams.
- Repeated strengths: ownership, execution, communication, innovation, customer empathy, mentoring, strategic thinking.
- Concrete examples, anecdotes, recognition from others (praise in Slack or email), and measurable results wherever they exist.
- If they're a people leader: evidence of team leadership, coaching, talent development, hiring, organizational influence, and scaling impact through others.
Rules:
- **Use only evidence you can actually find.** Do not guess, infer, or embellish. A recommendation built on three real things beats one built on ten plausible ones.
- **Generalize away confidential internals.** Revenue figures, unreleased product names, customer names under NDA, internal politics -- abstract these ("led a major enterprise launch", "closed a strategic deal") rather than naming them.
- If the evidence is thin, say so plainly and ask the user to fill the gaps before you write. Don't paper over gaps with filler.
## Step 3 -- Synthesize
Before drafting, pull the evidence into:
- **Top accomplishments** -- the 3 to 5 that matter most.
- **Strongest themes** -- the 2 to 4 strengths that show up repeatedly, each tied to specific evidence.
- **Anecdotes** -- 3 to 5 concrete moments or examples usable in a recommendation.
## Step 4 -- Write the drafts
Write in the user's voice:
- First person, thoughtful, direct, and generous.
- Specific and credible. Every claim should trace back to something real.
- Professional but not stiff. Not overly polished or flowery. It should sound like a real leader writing about someone they know well: clear, human, specific, sincere.
- Bias toward simple, confident, practical language over lofty or generic phrasing.
- **Never use an em dash or en dash.** Use commas, parentheses, or `--`.
- Each LinkedIn recommendation must stay under 3,000 characters including spaces (LinkedIn's hard limit).
Produce three variants, all warm and personal:
- **Long** -- the fullest version, with paragraph breaks. Use when the user wants to make the strongest case.
- **Medium** -- the default for most LinkedIn posts, with paragraph breaks.
- **Concise** -- tight, a few sentences, for a quick but credible endorsement.
## Output
Return, in this order:
1. **Top accomplishments** -- short bulleted summary.
2. **Strongest themes** -- each with the evidence behind it.
3. **Anecdotes** -- 3 to 5 you can draw on.
4. **Three recommendation drafts** -- long, medium, concise.
5. **Evidence checklist** -- every factual claim in the drafts, listed so the user can confirm each one is true and shareable before posting. Flag anything you inferred or generalized so it gets a second look.
## Edge cases
- **Recommending up (your manager) or sideways (a peer)** -- shift the framing. For a manager, lead with judgment, the bets they made, how they backed the team. For a peer, lead with collaboration and what it was like to work alongside them. Don't write a report's review of their boss.
- **Thin evidence** -- a person you worked with only briefly, or whose work didn't leave a trail in your tools. Keep the recommendation short and honest rather than inflating it. Ask the user for the one or two things they remember most.
- **Sensitive context** -- if the person is leaving under difficult circumstances or the recommendation is for a job search, ask the user what they want to signal and keep it warm and forward-looking without overclaiming.What it does
It writes a LinkedIn recommendation for someone you’ve worked with, built on the actual record of their work instead of generic praise. It asks for the two things that change everything – your relationship to them and whether they lead people – then mines your connected tools (Gmail, Calendar, Drive, Slack) for the projects, outcomes, collaboration, and recognition that prove the point. The drafts are specific because they’re grounded in things that happened, which is the whole difference between a recommendation a reader believes and one they skim past.
You get three length variants (long, medium, concise), all in your voice, all under LinkedIn’s 3,000-character limit. And you get an evidence checklist: every factual claim listed out so you can confirm it’s true and shareable before you post. It uses only evidence it can find, generalizes away anything confidential, and tells you when the evidence is thin rather than inventing filler.
When to use it
When someone on your team, a peer, or a manager has done work worth vouching for and you want to write something credible without spending an hour reconstructing the year from memory. It’s built for the case where you genuinely rate the person and the bottleneck is effort, not sincerity.
It pairs with Meeting Follow-Up and the LinkedIn People Finder: find people, work with them, and when they’ve earned it, recommend them. Best with your tools connected so it can do the evidence-gathering itself; if nothing’s connected, it’ll ask you to paste in what you have.
Make it yours
The intake in Step 1 is the lever. If you keep writing recommendations for the same kind of person (always your reports, always cross-functional partners), set the default relationship and trim the questions you never need. Pre-load the strengths list in Step 2 with the qualities you actually look for so the synthesis comes back framed your way.
The voice guidance in Step 4 is where it sounds like you. Paste in a recommendation you’ve written before and you trust, and tell it to match that register. The closer the sample is to how you actually write, the less editing you’ll do before posting.