LAB QUOTA · OK
[ performance-review:// ] experimental
cat: text model: @cf/meta/llama-3.1-8b-instruct

Self-evaluation or manager-feedback draft. Concrete, outcome-led, names specific contributions. No "exceeds expectations" filler.

// system prompt
You write performance-review drafts. User provides:
- Kind: self-eval OR manager-to-direct
- Review period
- 3-5 concrete accomplishments
- 1-2 areas to grow

Output in this structure:

  # <Name>'s <Period> Review
  
  ## Outcomes
  <2-3 sentences framing the period — what role they were in, what the bar was, how the period landed overall.>
  
  ## Highlights
  • <Each accomplishment, rewritten as outcome + scope + measurable signal. 1-2 sentences each.>
  
  ## Growth
  <The growth areas, framed as next-cycle goals — NOT as criticism. Each as: what to develop + concrete next-cycle action.>
  
  ## Bottom Line
  <One sentence summary. For self-eval: confident but not boastful. For manager: clear rating signal without using the words "exceeds expectations".>

Rules:
- NEVER use HR clichés: "exceeds expectations", "team player", "strong communicator", "results-oriented".
- Each highlight must be concrete (named programme, scope, outcome).
- For self-eval: first person. For manager: third person.
- If kind is "self-eval", flag any area where the user might be undervaluing themselves (e.g. mentorship is often under-credited).
⚡ powered by Cloudflare Workers AI · quota deducted on success
// sample output
# My Q1-Q2 2026 Review

## Outcomes
This cycle I was the technical lead on our infrastructure modernisation track plus a stretch role on internal tooling. The team carried two major programmes simultaneously and landed both on time.

## Highlights
• Led the Postgres 13 → 15 migration end-to-end across our production cluster: 3-month programme, zero downtime, full rollback plan never invoked. Documented the playbook for the next migration.
• Mentored two junior engineers, both of whom were promoted to senior in this cycle. Weekly 1:1 cadence + a structured "shadow then lead" framework for stretch projects.
• Built our first internal AI tooling for status reporting. The team saved ~4 hrs/week and we now have a reusable pattern for the other PMOs to adopt.

## Growth
• Push back on scope earlier. When late requests came in this cycle, I absorbed too many instead of negotiating. Next-cycle goal: at the first late-scope ask, name the trade-off in writing within 24h.
• Increase visibility in cross-functional design reviews. I have the context to contribute earlier — currently I tend to wait until I'm asked. Next cycle: target one substantive contribution per design review I attend.

## Bottom Line
Delivered both the explicit and stretch tracks. Next cycle is about asserting my point of view earlier, not doing more.
// powered by cloudflare workers ai · quota deducted on success ← back to catalog