📊 Agentic Workflow Lock File Statistics - February 4, 2026 #13689
Closed
Replies: 1 comment
-
|
This discussion was automatically closed because it expired on 2026-02-11T08:32:23.349Z.
|
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Executive Summary
This comprehensive analysis examines 145 lock files totaling 9.19 MB in the
.github/workflows/directory, revealing patterns in trigger configurations, safe outputs, structural complexity, and engine distribution across the repository's agentic workflows.Key Highlights:
📁 File Size Distribution
Statistics:
View Top 5 Largest Workflows
🎯 Trigger Analysis
Most Popular Triggers
Common Trigger Combinations
Insight: 64.8% of workflows follow the "scheduled with manual override" pattern, enabling both automated periodic execution and on-demand testing/debugging.
Schedule Patterns
Top Scheduled Times (Weekday Daily Runs):
View Schedule Distribution
0 14 * * 1-50 13 * * 1-50 11 * * 1-50 9 * * 1-50 7 * * 1-5*/4,*/6,*/12)Pattern: Most workflows run during business hours UTC (9 AM - 4 PM), concentrated around early afternoon. This likely aligns with development team working hours for result visibility.
🔒 Safe Outputs Analysis
Safe outputs enable workflows to create discussions, issues, comments, and pull requests in a controlled manner.
Safe Output Types Distribution
Key Observations:
create-discussion- the dominant output mechanismExample Workflows by Safe Output Type
create-discussion workflows:
create-issue workflows:
create-pull-request workflows:
add-comment workflows:
🏗️ Structural Characteristics
Job Complexity
Job Distribution Pattern:
Most workflows follow a 6-job structure:
Top 5 Most Complex Workflows (by step count)
These complex workflows typically involve multi-stage data processing, external API calls, and comprehensive reporting.
Average Lock File Structure
Based on statistical analysis, a typical .lock.yml file has:
🔐 Permission Patterns
Workflows primarily request minimal, job-specific permissions following the principle of least privilege.
Common Permission Sets:
Permission Strategy:
⚙️ Engine & Tool Distribution
Engine Distribution
Note: Total > 145 indicates some workflows may test multiple engines or have engine-switching capabilities.
Observation: Copilot dominates the engine distribution, powering over 70% of workflows. Claude is the second choice for nearly 30% of workflows.
Common MCP Servers & Tools
Based on docker image and server configurations:
Most Common Base Images:
node:lts-alpine- 118 instancesv0.30.3- 156 instancesSpecialized MCP Servers Detected:
⏱️ Timeout Patterns
Timeout Distribution:
🔍 Interesting Findings
1. Discussion-First Culture
94.5% of workflows output to discussions rather than issues, indicating a preference for conversational, less formal output for reports and analyses. Issues are reserved for actionable items requiring tracking.
2. Schedule + Manual Override Pattern Dominance
Nearly 65% of workflows use the
schedule + workflow_dispatchtrigger combination, showing a mature workflow design that balances automation with flexibility for testing and debugging.3. Consistent 6-Job Architecture
Most workflows follow a standardized 6-job pattern (pre_activation → detection → activation → agent → safe_outputs → conclusion), demonstrating strong architectural consistency and potentially shared templates.
4. Copilot Adoption
With 71% of workflows using Copilot as the engine, there's clear organizational preference, though Claude (28.3%) and Codex (9%) maintain significant presence for specific use cases.
5. Afternoon UTC Scheduling Bias
Scheduled workflows cluster around 11 AM - 2 PM UTC (weekdays), suggesting optimization for visibility during primary development team working hours.
6. Size Consistency
91.7% of lock files fall within 50-100 KB range, indicating standardized workflow complexity. The few outliers (< 50 KB or > 100 KB) represent either simplified test workflows or highly complex multi-stage pipelines.
7. Agent Job Complexity
The "agent" job typically contains 30-55 steps and represents the core agentic work. This job is significantly more complex than supporting jobs (3-12 steps), showing clear separation of concerns.
8. Multi-Modal Safe Outputs
75% of workflows use multiple safe output types, enabling rich, multi-channel communication (e.g., create discussion for summary + create issue for action items).
📈 Historical Context
Current Snapshot (2026-02-04):
This represents a mature, production-scale agentic workflow repository with strong architectural patterns and operational practices.
Note: Historical trend data will be available in future analyses as multiple snapshots accumulate.
💡 Recommendations
1. Optimize Large Workflows
The top 5 workflows with 45-55 steps could benefit from modularization review. Consider breaking complex agent jobs into sub-workflows or reusable actions.
2. Standardize Timeout Values
With median at 15 minutes but average at 16.58 minutes, consider standardizing to 15 or 20-minute increments for consistency unless specific requirements differ.
3. Engine Strategy Documentation
With 3 primary engines in use (Copilot 71%, Claude 28%, Codex 9%), document the selection criteria for when to use each engine to guide future workflow development.
4. Schedule Distribution
Consider distributing scheduled workflows more evenly across the day (currently clustered 11 AM-2 PM UTC) to reduce potential resource contention and spread system load.
5. Safe Output Consolidation
With 94.5% using discussions, consider formalizing discussion categories and naming conventions to improve discoverability of automated reports.
6. Test Workflow Cleanup
Review the 10 smallest workflows (10-50 KB range) to determine if they are still actively used or can be archived/consolidated.
🛠️ Methodology
Analysis Tools
/tmp/gh-aw/cache-memory/Data Sources
.github/workflows/*.lock.yml(145 files)Validation
ls -land stat commandsCache Memory Structure
Analysis scripts, historical data, and extraction patterns stored in
/tmp/gh-aw/cache-memory/for future reuse and trend analysis:history/2026-02-04-analysis.json- Complete statistical snapshotscripts/analyze_lockfiles.sh- Reusable bash analysis scriptpatterns/quick_commands.sh- Useful one-liner commandsREADME.md- Cache documentationReferences:
Generated by Lockfile Statistics Analysis Agent on 2026-02-04
Beta Was this translation helpful? Give feedback.
All reactions