Back to use cases
People & HR

Draft a performance review from the work itself

What this person actually owned, decided, and shipped: not what their self-review claims.

Best for
Calibration season
Primitives
ActorsDecisionsOwned Work
Token savings
63%

The prompt

You are drafting a performance review for {{person_name}} for the period {{start_date}} → {{end_date}}. Respect the access scope of the person running this prompt — the Sentra envelope will surface a "withheld" flag for anything they can't see; do not invent around it. Using Sentra: 1. Pull every Decision {{person_name}} made or led in the period, with Rationale. 2. Pull every Commitment they owned, and the delivery status. 3. Pull collaboration signal — who they worked with most often, and how that work was received (sentiment of Interactions where they spoke). 4. Pull any moment Sentra flagged where they were named as a blocker or as a champion by someone else. 5. Pull their own self-described areas of focus from any career-conversation Interactions (1:1s, growth docs). Output a draft review with these sections: - Impact: top three pieces of owned work with evidence. - Collaboration: how they show up cross-team, with two specific examples. - Growth: where Sentra has seen them stretch, and where they've stayed inside a comfortable surface. - Forward: three concrete asks for the next period. End with a one-line readout of source coverage so the manager knows what's grounded and what isn't.

See it work

Run it from

ClaudeCursorChatGPTAny agent · MCP

How it runs

  1. 01Copy the prompt into any agent connected to Sentra over MCP.
  2. 02The agent reads the company memory, scoped to what you are allowed to see.
  3. 03One grounded answer comes back, with every line traceable to its source.

Or automate it with Actions

The same prompt runs as a Sentra Action: on a schedule, or on a semantic trigger. Output lands in Slack or email, and every send waits for your approval.

Trigger idea: calibration season

Sentraover MCPFrom the work itself, every line traceable

Devon · H1 review draft

Owned

8 sources

Billing reliability end to end, including the invoice-retry redesign that closed the incident cluster.

Decided

design review · Aug 8

Argued for and won the retry-queue split; introduced revert discipline after the Sep 3 incident.

Cross-team

threads · Aug

Unblocked the Solara SSO audit twice without being asked; both saves are in the account thread.

Growth edge

Design docs land late relative to build starts; three of four this half were written after the first PR.

The prompt above becomes this. Fictional data, real mechanics: every line cites the meeting, thread, or record it came from.

Without Sentra

  1. 01Pull every meeting transcript, thread, email, and 1:1 note involving the person over the period.
  2. 02Read each to identify decisions led, commitments owned, collaboration moments.
  3. 03Re-read to extract sentiment of interactions where they spoke — was the work well-received?
  4. 04Search for moments where the person was named as a blocker or as a champion by others.
  5. 05Read their self-described areas of focus from 1:1 notes and growth docs.
  6. 06Compose the review with inline citations.

~140,000 tokens

With Sentra

  1. 01Pull every `Decision` where `Actor` = person, with `Rationale`, for the period.
  2. 02Pull every `Commitment` they owned and the delivery status.
  3. 03Pull collaboration signal — who they worked with most, with sentiment of `Interactions` where they spoke.
  4. 04Pull moments where they were named as blocker or champion by others.
  5. 05Pull career-conversation `Interactions` for self-described focus areas.
  6. 06Compose the review with source coverage readout.

~52,000 tokens

Book a DemoStart Free