Performance-Review Season Without the Dread
Turn a year of scattered notes into fair, specific reviews — self-assessment prompts, manager drafts grounded in real examples, and calibration prep that holds up.
A review is only useful if it points at something specific the person can act on.
The dread comes from facing a blank box and a year you half-remember. Store your standard — your competencies, your rating scale, the examples you've jotted along the way — and the review becomes assembling evidence you already have instead of inventing an assessment under deadline.
Vague reviews protect no one. 'Great team player, keep it up' tells someone nothing, and 'needs to improve communication' with no example is something they can't fix.
Feedback without a specific example is an opinion. Feedback with one is something a person can actually use.
Capture examples through the year in a running file, not from memory in December. The review writes itself from notes you already took.
| Complexity | Easy |
| Tools needed | Claude Pro or above, Claude Desktop → Cowork mode, on macOS or Windows, Claude Projects |
| Time to build | ~25 min first time · ~10 min per person after that |
| Best for | Manager or founder writing reviews for a direct team — B2B services and small companies |
What this solves
Review season spent reconstructing a year from memory and writing safe, vague feedback nobody can act on.
The problem
It’s review season and you have six to write. You sit down to do the first one and realize you can remember the last two months clearly and the ten before that as a blur. So you write something safe and general — “strong contributor, great attitude” — which helps no one, or you reach for the one incident you do remember, which makes the whole review feel like it’s about that. The person leaves the conversation without a clear picture of where they stand or what to do next. That’s not a rating problem. It’s an evidence problem.
The work is gathering the specifics. Do that, and the review almost writes itself — fairly.
Ingredients
- Claude Subscriptions — Cowork and Projects aren't available on the free plan
- Platforms & Modes — Cowork runs in the desktop app only — not on web or mobile. This recipe pulls together notes and produces structured review drafts — multi-step work across files
- Claude Projects — Your review framework — competencies, rating scale, tone — is the same every cycle. A Project holds it so every review is built the same way and stays consistent across your team
How it works
Reviews repeat every cycle and span your whole team. A Project stores your review framework — the competencies, the rating scale, your tone — so every review is built the same way and you’re not re-establishing the standard for each person.
- Open Claude Desktop and click Cowork in the mode selector across the top (Chat · Cowork · Code).
- In the left panel, find Projects and click the + button.
- Choose Start from scratch. Name the project “Performance Reviews” and let Claude set up its folder.
- You’ll know it worked when the project appears in the left panel with its own folder and an instructions field.
Your project has a folder on your computer — that’s where the working files live: review-framework.md plus one running notes file per person.
- Ask Claude, right in the project: “Create
review-framework.mdin this project’s folder, plus one notes file per person on my team —notes-jordan.md,notes-sam.md, and so on. Leave them empty — I’ll fill them in.” (Or create them yourself in any text editor and save them into the project folder.) - Fill them in using the descriptions below.
- Confirm Claude can see them: ask “List the files you can see in this project.” The framework and every notes file should come back. If they don’t, see If It Doesn’t Work.
review-framework.md
Your standard: the competencies you review on, your rating scale with anchors, and the tone you want (direct, specific, forward-looking). The Knowledge File Seeding guide helps if you’re starting from scratch.
Example: “Competencies: impact, collaboration, ownership, growth. Ratings: exceeds / meets / developing, each with a one-line definition. Tone: specific and kind. Every rating must cite at least one concrete example. No rating changes in calibration without new evidence.”
notes-[name].md
A running file you add to through the year — wins, misses, moments worth remembering, dated. This is the file that ends the December scramble. Even three lines a month makes review season trivial.
Example: “Mar 4 — ran the vendor migration solo, caught a data issue before launch. May — missed two deadlines on the reporting project, root cause was over-committing, not capability. Sept — mentored the new hire, unprompted.”
Open your Project in Claude Cowork. Give Claude the specifics for this run, ask for the main output, then follow up for any additional pieces you want. The exact wording for each prompt — with what it’s asking for and why — is in What you actually type below.
Check three things before this becomes a conversation:
- Every rating cites a real example. If a rating leans on a general impression instead of a moment from the notes, that’s the rating to either find evidence for or soften. A review the person can’t connect to anything they did is one they can’t act on.
- The gaps are flagged, not filled. Where you don’t have evidence, the draft should say so — not manufacture a plausible-sounding example. Confirm Claude flagged thin areas instead of smoothing over them. Inventing evidence is the one thing that breaks a review’s trust.
- The development points are specific and doable. “Improve communication” is useless. “Send a written recap after client calls so context doesn’t get lost” is something Jordan can start Monday. Push vague growth areas into concrete next actions.
Before the conversation: does this match what you actually believe about this person? Claude assembled the evidence; the judgment is yours. If the draft’s conclusion feels off, the notes are incomplete — trust your read and add the missing context.
What you actually type
Name your files explicitly the first few runs, and ask Claude to show its work on anything that matters.
“Using `review-framework.md` and `notes-jordan.md`, draft Jordan's review. For each competency, give a rating and back it with a specific example from the notes. Flag any competency where I don't have enough evidence to rate fairly.”
“Draft three self-assessment questions to send Jordan before our conversation, matched to these competencies.”
“Summarize the case for an 'exceeds' on ownership in two sentences I can defend in calibration.”
A review draft with a rating for each competency in your framework, every rating backed by a dated example from the person's notes file, plus development points written as concrete next actions. Competencies without enough evidence are flagged as unratable, not filled in. If a rating leans on a general impression, or a thin area got a confident rating instead of a flag, it missed — send it back.
- It doesn't replace your judgment. Claude assembles the evidence; whether the conclusion matches what you believe about this person is your call to make.
- It's only as good as the notes. A file with two entries produces a review with two examples — the recipe's real work is the three lines a month you add all year.
- It doesn't deliver the review. The conversation, the reaction, and what the person does next are still yours to handle.
If it doesn’t work
- No Cowork tab in Claude Desktop — update the app to the latest version and confirm you’re on a paid plan; Cowork isn’t on the free tier. On Windows, Cowork also needs the Virtual Machine Platform feature enabled — if the tab still won’t appear, that’s the fix.
- Claude can’t see
review-framework.mdor the notes files — the files aren’t in the project’s folder, or they’re in a different folder than the one the project owns. Open the project, check which folder it points to, and move the files there. Then re-run “list the files you can see.” - A rating leans on a general impression, not a moment — the notes file is thin. Two entries for the year gives Claude two examples to work with, and the rest comes out as filler. Add the moments you do remember — dated, one line each is enough — and re-run. Then set the monthly nudge in Extra Credit so next cycle isn’t archaeology.
- An example shows up that isn’t in the notes — the draft is filling gaps instead of flagging them. The prompt’s last line does the guarding: “Flag any competency where I don’t have enough evidence to rate fairly.” If something still looks invented, ask: “Which note is each example from? Remove anything you can’t trace to the file.”
Extra credit
Small additions that pay back the next time you run it.
- Running notes all year — set a scheduled task that nudges you on the first of each month to add three lines to each person’s notes file. Review season stops being a season.
- Google Drive — keep the framework in a shared Drive folder so every manager reviews on the same standard. See the Connectors guide.
- Calibration prep — paste all your drafts in and ask Claude to surface rating-distribution and where two people with similar evidence got different ratings, before the calibration meeting.
“The hardest part of a review isn't the rating. It's remembering the specific moment that justifies it.”
What this teaches you about Claude Cowork
The recipe is one application. The principles apply to everything you’d hand to Claude.
The evidence problem is the real problem. Reviews dread-inducing because you’re reconstructing a year from memory. A running notes file moves the work to where it’s painless — three lines a month — and makes the review an assembly job instead of an act of recall.
Flagging beats filling. The valuable behavior is Claude telling you where the evidence is thin, not papering over it. A review built only on what actually happened is one you can stand behind in calibration and in the room.
Specific is kind. Vague feedback feels safe but helps no one. The discipline of citing an example for every rating — enforced by the file — is what turns a review from an ordeal into something the person can actually use to get better.
Who this is for
Manager or founder writing reviews for a direct team in B2B services and small companies (5–50 employees).
The pain: Six reviews due and only the last two months in clear memory
The outcome: Every review cites real examples and holds up in calibration without backpedaling
Published June 10, 2026 · 0 views