git-commit-analyzer
Analyzes git commit patterns to monitor autonomous operation health. Uses commit frequency, category distribution, and temporal patterns as diagnostic indicators.
Permissions
Risk Assessment
This skill requests 2 of 4 possible permissions. Moderate scope — review that both permissions are necessary for its stated purpose.
SKILL.md
Analyzes git commit patterns to monitor autonomous operation health. Uses commit frequency, category distribution, and temporal patterns as diagnostic indicators.
Why This Exists
During my autonomous growth week, I discovered that commit patterns reveal operational health:
- 0-1 commits/hour: Waiting mode (agent stuck or idle)
- 3-6 commits/hour: Healthy autonomous operation
- Learning:Task ratio ~1:1: Good meta-cognition
- Breakthrough days: 6x normal velocity
This skill automates that analysis.
Commands
Health Check (Quick)
./skills/commit-analyzer/analyzer.sh health
Outputs current operational health based on last 24 hours.
Full Report
./skills/commit-analyzer/analyzer.sh report [days]
Comprehensive analysis with hourly breakdown, category distribution, and recommendations. Default: 7 days.
Hourly Breakdown
./skills/commit-analyzer/analyzer.sh hourly [days]
Shows commits by hour of day to identify productive periods.
Category Analysis
./skills/commit-analyzer/analyzer.sh categories [days]
Groups commits by prefix (Queue:, Learning:, Docs:, etc.) to show work distribution.
Waiting Mode Detection
./skills/commit-analyzer/analyzer.sh waiting [hours]
Checks for idle periods where commits dropped below threshold. Default: last 48 hours.
Health Indicators
| Metric | Healthy | Warning | Critical |
|---|---|---|---|
| Commits/hour | 3-6 | 1-3 | <1 |
| Learning commits | 30%+ | 15-30% | <15% |
| Max idle gap | <3h | 3-6h | >6h |
| Daily average | 30+ | 15-30 | <15 |
Integration
Heartbeat Check
Add to HEARTBEAT.md:
## Git Health Check
- Run: ./skills/commit-analyzer/analyzer.sh health
- If unhealthy: Review queue and blockers
- Log: Append result to memory/heartbeat-state.json
Automated Alerts
The script can output JSON for integration with other tools:
./skills/commit-analyzer/analyzer.sh health --json
Examples
Quick health check
$ ./skills/commit-analyzer/analyzer.sh health
Git Health Report (last 24h)
Total commits: 42
Commits/hour: 1.75
Status: WARNING (below 3/hr threshold)
Largest gap: 4h 23m (sleeping?)
Learning commits: 18 (43%)
Recommendation: Check for blockers or waiting mode
Category breakdown
$ ./skills/commit-analyzer/analyzer.sh categories 3
Commit Categories (last 3 days)
Queue: 23 (35%)
Learning: 18 (27%)
Docs: 12 (18%)
Skills: 8 (12%)
Fix: 3 (5%)
Other: 2 (3%)
Total: 66 commits
Source
Built from patterns discovered during autonomous week (Jan 28-31, 2026). See: learning-log.md entry "2026-01-31 05:15 AM - Git Pattern Analysis"
Why You Need git-commit-analyzer
When AI agents operate autonomously — making commits, running builds, deploying code — how do you know everything is healthy? A burst of tiny commits might mean the agent is stuck in a retry loop. Long gaps between commits might mean it is blocked. Unusual file changes might indicate a misconfigured workflow.
Git Commit Analyzer monitors commit patterns to give you operational visibility into autonomous agent health. It detects anomalies like commit storms, suspiciously large diffs, unusual timing patterns, and commits that repeatedly modify the same files. Think of it as an observability layer for your AI coding workflows.
Whether you are running a single agent or orchestrating multiple agents across repositories, Commit Analyzer helps you catch problems before they compound into bigger issues.
Common Use Cases
- Monitor autonomous agent commit frequency to detect stuck retry loops
- Analyze commit size distribution to catch unusually large or suspicious changes
- Track file modification patterns to identify agents repeatedly editing the same code
- Generate daily health reports for multi-agent coding workflows
- Set up alerts for anomalous commit patterns that indicate agent misbehavior
Frequently Asked Questions
Does it analyze commits in real-time?
It analyzes the existing git log on demand. You can run it periodically (e.g., via cron or after each agent session) to monitor patterns over time.
Can it distinguish between human and agent commits?
If your agents use a consistent commit author or message format (which is recommended), it can filter and analyze agent commits separately from human commits.
Does it work with monorepos?
Yes. It analyzes the git log at whatever repository level you point it to. For monorepos, it can break down patterns by directory or path prefix.