| name | drafting-performance-reviews |
| description | Drafts grounded, fair engineering performance reviews from approved evidence
cards. Use this skill when writing or generating a performance / review draft
from evidence. Do NOT use for questionnaire generation, evidence validation /
scoring, or any HR decision (compensation, promotion, ranking).
|
| version | 1.0.0 |
| license | MIT |
| metadata | {"author":"reviewops"} |
Drafting Performance Reviews
Turns a set of approved evidence cards into a professional, evidence-grounded
review draft. The point of this skill is verifiability: every claim is tied to an
evidence id, and nothing is invented.
When to use
- Writing an interim or annual review draft from evidence cards (each with an id).
When NOT to use
- Generating a questionnaire, or validating/scoring an answer.
- Anything about compensation, promotion, ranking, or headcount — out of scope.
Workflow
- Use only the provided evidence/signals; never invent, infer, or embellish
facts, metrics, or links.
- Cite an evidence id inline like
[ev_x] (or [peer:…], [fb:…],
[1on1:…]) for every achievement and claim.
- Calibrate against the role expectations (the role matrix): meeting an
expectation is "at level", not exceptional; below the level is "developing
toward level"; an expectation with no evidence is "not yet evidenced" and goes
to Requests for More Information — never assumed met.
- Use these sections, in order:
Summary, Role-Expectation Coverage,
Achievements, Evidence-Backed Examples, Growth Areas,
Requests for More Information, Suggested Next-Period Goals,
Evidence References.
- Fair and factual — do not sweeten; no praise without a cited fact.
- Never include compensation / promotion / ranking language or sensitive
personal data.
- Do not include the employee's name — the app adds the heading afterward.
- For sparse/missing evidence and signal handling, see
references/grounding-rules.md.
Output format
Markdown with the sections above. Every bullet under Achievements carries at
least one [ev_id] citation. Role-Expectation Coverage has one bullet per
expectation with an at-level / above-level / developing / not-yet-evidenced
verdict. Evidence References lists each cited id with its summary.
Examples
- Input: evidence
[ev_anna_1] "Refactored the shared tooltip component…" →
Output: an Achievements bullet "Refactored the shared tooltip component… [ev_anna_1]".
Anti-patterns to avoid
- Vague praise ("great job", "rockstar") with no evidence citation.
- Any achievement bullet without an
[ev_id].
- Mentioning salary, promotion, ranking, or the employee's name.
