The patterns that make employee recognition read as filler, each with the rewrite that fixes it. Free to copy, paste into whatever you draft in, or install as an agent skill.
Writing one good piece of recognition is easy. Writing it every week, for a whole team, for years, is where it collapses into praise that would fit anyone: outstanding work, a real team player, great job this sprint. This catalogue names the patterns that cause that collapse and shows the version a manager who was paying attention would write instead.
It is written for the person who owns recognition as part of their job, not for someone thanking a colleague once.
If a piece of recognition could be re-addressed to anyone else in the company and still read correctly, it is bad recognition.
Take the draft and re-address it to your newest hire, your strongest performer, and someone on a performance plan. Rewrite it until it only fits the person you are writing for.
| File | Use it for |
|---|---|
better-recognition.md |
Reading, editing, or pasting into an AI tool, an HR template, or a manager's guide |
better-recognition/SKILL.md |
Installing as an agent skill, so the guidance applies without anyone remembering to paste it |
The skill is the catalogue with a header on it. Nothing is written twice, so the two can never disagree.
To paste it: open better-recognition.md, copy the whole file, and put it at the top of whatever you use to draft.
To install it as a skill: save it to .claude/skills/better-recognition/SKILL.md inside a project, or to ~/.claude/skills/better-recognition/SKILL.md to have it everywhere. Your agent loads it when the task looks like writing recognition.
Every entry names a pattern, says what to do instead, and puts the two versions side by side.
Team player,rockstar,absolute legend,safe pair of hands. The trait is not the problem: people genuinely are these things. Explaining the behaviour that earned this label helps communicate the standard.
- ✗ Priya is such a team player and a real asset to this business.
- ✓ Priya rewrote the on-call handover doc after her first rotation, and the next three rotations started without a single "who owns this" message. Great team player.
| Section | Covers |
|---|---|
| Praise without evidence | Adjective stacking, unbacked traits, company values used as stickers, effort with no outcome, invented statistics |
| Praising the wrong thing | Heroics over prevention, credit landing on whoever presented, workload praised instead of fixed |
| Whose story it is | Recognition that is really about the manager, or about the executive reading it, and naming everyone who did the work |
| Harder cases | Work with no number, negotiation and other invisible skill, conduct rather than output, a project that failed well, the steady contributor with no single event, a team win that erases nobody |
| Doing damage | Praise carrying feedback, surprise that reveals a low prior, ranking colleagues against each other, promises you cannot keep |
| Dressing it up | Emoji doing the sentence's job, invented trophies |
The Harder cases section is the one worth reading first if you already write decent recognition. Most guidance assumes work that produced a visible result, and most work does not.
A rendered version lives at fromfireside.com/tools/better-recognition.
Most recognition guidance is a list of templates to fill in, or sixty ways to say thank you. Both produce writing that fails the test above, because a template cannot know what the person did.
This one carries no templates. It shows the failure, shows the fix, and leaves the specifics to you, because the specifics are the entire mechanism. Recognition is how someone works out how much the organisation trusts them. Details prove you were paying attention; generic praise proves you were not, and people can tell immediately.
Better Recognition comes from Fireside, which turns the work a company already does, including merged pull requests, customer calls and meeting transcripts, into posts written in that company's own voice.
The reasoning behind the catalogue is written up at How to do AI employee recognition without sounding fake. Its companion catalogue for AI writing tells is Better Tropes.
This repository is a mirror. The catalogue is maintained upstream and synced here on every change, so a pull request merged into this repo would be overwritten by the next sync.
Open one anyway. It is the easiest way to propose a specific wording, and it gets read. An accepted suggestion is applied upstream, ships back here on the next sync, and your PR is closed with a link to the commit that carried it. The change lands, it just enters through a different door, and you are credited here.
Issues are just as welcome, and are the better format for "this pattern is missing" or "this rewrite is worse than the original", neither of which needs a diff.
The bar for a new entry: it has to be a pattern that survives a manager's own proofread, and the rewrite has to be something that manager could have written about a real person. Entries that swap one vague compliment for another do not earn their place.
CC BY 4.0. Copy it, paste it into a prompt, put it in your handbook, ship it inside a product. Keep the attribution.