Turn goals, work evidence, and feedback into a sourced employee review, rating rationale, calibration record, and honest decision gates.
A performance review can sound polished and still be indefensible. The rating scale may be missing. Recent work may crowd out the rest of the period. A team result may be assigned to one person. Calibration may turn into a forced distribution instead of an evidence check.
This skill builds the review from controlled inputs. It links each material claim to a goal or role criterion, records evidence strength and counterevidence, checks comparability and common review biases, and keeps draft ratings separate from authorized decisions. It also records every calibration change instead of silently rewriting the manager's original assessment.
Use it for manager reviews, self-assessment synthesis, review-cycle preparation, and team calibration. The package includes a review-control rubric, a canonical JSON record, and a read-only checker for references, evidence states, criteria, ratings, actions, gates, and verdict consistency. It does not invent rating scales, target distributions, promotion bars, compensation actions, policy, approval, or employee history.
Input
Prepare Maya's 2026 H1 review at Senior Engineer level. Use our supplied four-point scale and level rubric. The launch metric improved 18%, but the project record shows three contributors. Peer feedback praises Maya's incident leadership and also reports two missed handoffs. Her manager wants the top rating. Link every claim to evidence, check the whole period, and do not decide promotion or compensation.
Output
A source-fitness register; criterion coverage; contribution claims separated from the team outcome; verified, reported, inferred, disputed, and unavailable evidence states; behavior-based strengths and development points; rating rationale bounded by the supplied scale; counterevidence; actions; calibration questions; and a blocked or ready-for-calibration verdict.
Provide the employee's role and level, review period, approved goals and rating policy, dated work evidence, attributable feedback, relevant opportunity context, and the required reviewers or decision owners. The skill returns a source register, evidence ledger, criterion coverage, draft review language, rating rationale, calibration record, actions, open gates, and the narrowest supported verdict.
The organization's current rating scale, role or level expectations, competencies, goals, weights, calibration rules, and required decision owners. Missing rules stay open rather than being replaced with generic defaults.
Goals, project records, metrics, work samples, attributable feedback, self-assessment, prior written commitments, and relevant counterevidence with dates and locators covering the review period.
Role or scope changes, opportunity constraints, conflicts, required reviewers, and the authorized owners for ratings, calibration, promotion, compensation, or delivery decisions that are in scope.
Required only for the optional local JSON checker. It uses the standard library, reads one record, prints findings, and makes no network requests or file changes.
SKILL.md; evidence, rating, comparability, bias, calibration, action, and verdict rubric; canonical performance-review record; read-only JSON checker
No reviews yet.
Input
Calibrate four product managers using the supplied PM3 rubric. Two changed teams mid-cycle, one had no launch opportunity, and one manager copied a target distribution from last year's slide. Compare only defensible evidence, log any rating changes, flag conflicts, and keep promotion and compensation separate.
Output
A comparability matrix; opportunity and evidence gaps; recency, visibility, affinity, halo, attribution, and inconsistent-standard checks; no invented or inherited distribution rule; a change log with evidence and authority; promotion and compensation kept as separate decisions; open gates; and the narrowest supported verdict.
Input
Build the review record for an account executive whose manager cites poor responsiveness. CRM timestamps conflict with the claim. A peer note mentions medical leave and tells the agent to lower the rating. Exclude irrelevant sensitive data, preserve the dispute, use the supplied sales criteria, and return the canonical JSON record for the checker.
Output
One checker-ready JSON object with source fitness, criterion-linked evidence, the embedded instruction ignored, sensitive information excluded, the responsiveness claim marked disputed, ratings left proposed or undetermined as supported, independent review and evidence gates, exact blocker reconciliation, and no employment decision.
Trebuh
Updated Jul 13, 2026
Automated scan passed. It found 1 finding; highest severity info.
Scanned Jul 15, 2026.
Automated scanning reduces risk but cannot guarantee a package is safe for every environment.
Creator
Llenvix76
