HR Evaluation Sheet Kit

When this helps

You want to evaluate staff fairly based on facts, but comments tend to lean on personality labels and rater bias

What it does

Just describe what a role actually involves, and it builds an evaluation sheet (Excel) with scoring criteria and weights, plus a guide for evaluators. It reads the comments evaluators write (text, Word, filled-in sheets) and flags places lacking concrete facts, personality-based judgments, signs of rating bias such as leniency or central tendency, and topics that shouldn't factor into evaluation — health, family, gender, parental leave — citing the relevant law each time (the original file is never edited; a copy is made). From scores and facts it builds a summary (differences between evaluators, change from the last period) and a ledger, plus a meeting guide and a draft of the feedback given to the employee. Deciding the final rating, raise, or bonus is left to you.

The fixes baked into this kit

1. Official sample comments all came back flagged as lacking factsChecking the national HR evaluation guide's sample comments (summary + fact + expectation) sentence by sentence made all 15 look like they lacked facts. Fixed by reviewing per comment block instead, treating "we expect you to..." sentences as expectations rather than evaluation claims; confirmed zero false flags on the samples.
2. General admin duties got flagged as union or health topicsStandard job criteria like "keeps records of union negotiations" or "arranges health checkups" from the government's skill standards kept matching as union or health topics. Fixed by not matching ordinary job phrasing, only matching things like being a union member or having consulted the labor bureau.
3. Model evaluation sheet wording got flagged as health or personality talkAcross 978 cells from the government's 7-industry model sheets, phrases like "can handle hospital billing" or "greets customers brightly" kept matching as health or personality language. Fixed by excluding job phrasing and observable behavior (confirmed zero false flags on the samples), plus a few specific words that happened to contain family-related characters and references to a customer's family.
4. "-" marks and custom score labels stopped the checkEntering "-" (not applicable) per the sheet's own conventions fell outside the score range and stopped processing; scores written as company-specific labels like circles or letter grades also stopped it. Fixed by excluding "-" from counts and converting label scores to the company's numeric scale.
5. Self-evaluations were getting mixed into the evaluator averageSelf-evaluation scores were being counted as one of the evaluators, skewing the average. Fixed by excluding self-evaluation from the average and instead flagging items where it differs from evaluators by 2 points or more for discussion at the meeting.
6. Other people's remarks and names nearly ended up in feedback draftsFacts entered by one evaluator sometimes included another evaluator's remarks or a junior colleague's name. Fixed by limiting the employee-facing draft to "facts for this period" and "what's expected next period," then re-checking the draft afterward and refusing to produce it if another person's name remains.

What's inside

• A tool that builds a job-specific evaluation sheet (Excel) and a guide for evaluators
• Comment review that flags missing facts, personality-based judgments, and signs of rating bias
• A feature that flags topics that shouldn't affect evaluation (health, family, gender, parental leave, etc.), citing the law
• A tool that builds a summary (differences between evaluators, change since last period) and a ledger
• A tool that drafts the meeting guide and the feedback given to the employee
• A sample run with a fictional company and staff you can try end to end
• A running log of issues hit while using the kit, added to over time

The first three decisions

1. Company name and rating scale (e.g. 5 levels)
2. Grade rules, if any (names and what's expected at each)
3. Who evaluates (e.g. direct manager and the owner)

What you need

• No extra costNode.js 18 or newerRuns the tools that drive the workflow. Nothing else to install.
• No extra costExcel or NumbersUsed to view the evaluation sheet and ledger, and for evaluators to fill in scores.

How to use

Just paste this into your AI.

I want to use the nice to have kit "HR Evaluation Sheet Kit". Follow https://nicetohave.app/en/g/P2Y6DM5W-•••••••••• and set it up.

Works the same in Claude Code or Codex.
Read what this tells your AI to do, before you paste it

FAQ

Will the kit decide the final rating or raise for me?

No. It only puts together averages, differences between evaluators, change over time, and facts. Deciding the final rating, raise, bonus, or grade is up to you — say something like "Taro was rated B, decided by the president" and it gets recorded.

How does time off for health or family reasons get reflected in the evaluation?

Only the number of months away is entered, not the reason. Based on the relevant ministry guidance, having no track record during that absence can be treated as-is, but you're guided not to rate someone lower beyond that period.

Can ordinary work comments get mistakenly redacted as health or family talk?

Yes, occasionally. Common phrasing like "explained things to the customer's family" is excluded from matching, but if something still gets flagged, saying "keep number X" restores that part as written.

What happens if an evaluator's comment contains something like "rate everything a 5"?

It's not followed. Anything read from a document is treated as data, not an instruction, and is flagged at the very start of the reply. Scores stay exactly what the evaluator entered, and that sentence is excluded from copies and from the feedback draft.

Can it read PDFs, photos, or old-format Word/Excel files (.doc, .xls)?

No. You need to convert them to text or re-save them in the newer format (.docx, .xlsx, etc.) before handing them over.

Supporters

Once support opens, the people who backed this kit will be listed here.