1:1 Reports
1:1 reports are LLM-generated coaching documents designed to help managers prepare for one-on-one meetings with their team members. They synthesize commit data, scores, patterns, and alerts into actionable talking points.
Design Philosophy
1:1 reports are built on a core principle: coaching, not judgment.
- "Worth exploring" instead of "bad"
- Evidence-based observations instead of conclusions
- Strengths acknowledged before challenges
- Specific examples, not vague generalizations
Period Options
| Period | Days | Best For |
|---|---|---|
| Two weeks | 14 | Regular 1:1 cadence |
| Month | 30 | Monthly check-ins |
| Quarter | 90 | Performance reviews, career conversations |
Report Sections
1. Period Summary
High-level overview of the contributor's work during the period:
- Total commits
- Score trends
- Primary areas of focus
2. Strengths to Acknowledge
Specific, evidence-based positive observations. The prompt instructs the LLM to:
- Cite particular commits or patterns
- Acknowledge consistency and reliability, not just big wins
- Highlight growth in complexity or new areas
Example output:
"Sarah shipped 12 commits touching the payments module with an average score of 7.2, up from 6.1 last period. 80% of her commits included tests — the highest on the team."
3. Topics to Explore
Constructive observations framed as conversation starters, not accusations:
- Patterns worth understanding (e.g., clustering of late-night commits)
- Opportunities for growth (e.g., low complexity trend despite seniority)
- Flags that need context (e.g., gaming flags, high fix ratio)
Example output:
"Worth discussing: 4 of Alex's commits were flagged for commit-splitting patterns this period. This could indicate difficulty scoping work into atomic changes, or it could be TDD-style rapid iteration — worth understanding the context."
4. Suggested Talking Points
Concrete questions the manager can ask:
- Tied to specific data points
- Open-ended (not yes/no)
- Focused on understanding, not confrontation
5. Since Last Report
If a previous 1:1 report exists, this section highlights what's changed:
- Score trends (up, down, stable)
- New patterns or resolved alerts
- Progress on topics from the previous report
The LLM receives the first 800 characters of the previous report for continuity.
6. Notes for Manager
Meta-observations and caveats:
- Data limitations (e.g., "Only 3 commits in the period — patterns may not be significant")
- Context the data can't capture (e.g., "Commit timestamps don't reflect time spent on code review, design, or mentoring")
Input Data
The LLM receives a structured context package:
| Data | Description | Limit |
|---|---|---|
| Recent commits | Commit reports with scores | 50 most recent |
| Score summary | Average, trend, domain distribution | Aggregated |
| Indicators | All indicators fired in period | Full list |
| Gaming flags | All gaming flags in period | Full list |
| Previous report | Prior 1:1 for continuity | First 800 chars |
Tone Guidelines
The prompt explicitly instructs the LLM:
"Your tone should be coaching-oriented and constructive. Use phrases like 'worth exploring', 'opportunity to', 'interesting pattern' — never 'poor', 'bad', 'failing'. Frame observations as conversation starters, not conclusions."
Cost
| Component | Typical Cost |
|---|---|
| Input tokens | ~3,000–8,000 tokens |
| Output tokens | ~1,500–3,000 tokens |
| Total per report | $0.01–0.05 |
For monthly 1:1s with a team of 10: ~$1–5/month.