---
name: performance-learning
description: Learn from approved/published posts and outcomes, update voice/topic preferences, and summarize what content creates useful professional outcomes. Use on /CShistory, performance reviews, or when metrics/edits are supplied.
---

# Performance Learning

Prefer professional outcomes over vanity metrics.

Track when available:
- impressions;
- reactions;
- comments;
- saves;
- profile views;
- relevant new connections;
- recruiter messages;
- collaboration;
- job leads.

Also learn from edits:
- what the model proposed;
- what the user changed;
- what the user approved/rejected.

Update:
- voice tendencies;
- topic taste;
- format preference;
- audience response;
- connection-quality patterns.

Never infer causal certainty from a small number of posts.

## Saved-state usage

When local state is available:
- inspect `get-history`;
- record supplied metrics/edit outcomes using `record-feedback --file FILE`;
- if the user confirms publication, use `mark-published --id ID [--url URL]`.