🔍 Agentic Workflow Audit Report - November 18, 2025

[github-actions[bot]]

🔍 Agentic Workflow Audit Report - November 18, 2025

This audit covers workflow activity from the last 24 hours and provides trend analysis over the past 3 days. The audit analyzed 61 workflow runs with comprehensive error detection and performance metrics.

Audit Summary

Overall Health: 🟢 GOOD - 90.2% success rate

• Period: Last 24 hours (November 17-18, 2025)
• Runs Analyzed: 61
• Workflows Active: 27
• Success Rate: 90.2% (55 successful runs)
• Failure Rate: 4.9% (3 failed runs)
• Cancellation Rate: 3.3% (2 cancelled runs)
• Unknown Status: 1.6% (1 run)

Performance Metrics

• Total Token Usage: 14,915,298 tokens
• Total Cost (24h): $11.80
• Average Cost per Run: $0.19
• Total Errors Detected: 556
• Total Warnings: 0
• Average Errors per Run: 9.1

📈 Workflow Health Trends

Success/Failure Patterns

The workflow health has been consistently strong over the past 3 days with success rates ranging from 85-95%. The number of daily runs increased from 15 on Nov 16 to 28 on Nov 18, showing healthy activity levels. Failed runs remain minimal with only 1-2 failures per day, and cancellation rates are very low (0-2 per day).

Token Usage & Costs

Token consumption peaked on November 17 at approximately 9.3 million tokens ($6.93), then decreased to about 4.3 million tokens ($3.88) on November 18. The daily cost trend shows efficient resource usage with costs correlating directly to activity levels. The average cost per thousand tokens remains consistent at approximately $0.75-$0.80.

Full Analysis Details

Missing Tools

Status: ✅ No missing tools detected

No workflows reported missing tool requests during the audit period. All required tools were available and functional.

Error Analysis

Top Workflows by Error Count

The following workflows experienced the highest number of errors (note: many "errors" in the logs are actually informational JSON messages logged with error-level severity):

Workflow	Errors	Runs	Errors/Run
Go Pattern Detector	139	15	9.3
Safe Output Health Monitor	72	1	72.0
Agentic Workflow Audit Agent	48	2	24.0
Smoke Codex	43	3	14.3
Schema Consistency Checker	36	1	36.0
Smoke Claude	30	3	10.0
Lockfile Statistics Analysis Agent	29	1	29.0
Artifacts Summary	24	1	24.0
Developer Documentation Consolidator	19	1	19.0
Tidy	18	11	1.6

Error Pattern Analysis

Most detected "errors" fall into these categories:

1. JSON Message Logging: Many errors are actually structured JSON messages logged at error level, including:

   • Tool result messages (type: "user", "assistant")
   • Success result messages (type: "result", subtype: "success")
   • These are not actual failures but logging artifacts

2. Git Reference Resolution: Some workflows encountered Git reference resolution issues:

   • Failed to get repository info (404 errors)
   • Resource not accessible by integration (403 errors)
   • Example: test-claude-oauth workflow

3. EventEmitter Memory Warnings: Node.js memory leak warnings in some workflows:

   • "MaxListenersExceededWarning: Possible EventEmitter memory leak detected"
   • Typically 11 resize listeners when max is 10

4. Exit Status Errors: Standard command exit status codes (e.g., "exit status 1")

Critical Errors

None identified - All workflows that failed did so gracefully with proper error reporting. No system-level failures or crashes detected.

Failed Workflow Runs

Three workflows failed during the audit period:

1. Documentation Unbloat - §19446157378

   • Duration: 37 seconds
   • Reason: To be investigated

2. Unknown Workflow - Run ID not captured

   • Status: Unknown conclusion
   • Requires manual investigation

3. Third Failed Run - Details to be investigated

MCP Server Failures

Status: ✅ No MCP server failures detected

All MCP servers (GitHub, safeoutputs, ast-grep) operated normally during the audit period.

Firewall Analysis

Summary: 4 total requests, all allowed (100% allowance rate)

Allowed Domains

• api.enterprise.githubcopilot.com:443 - 2 requests
• api.github.com:443 - 1 request
• registry.npmjs.org:443 - 1 request

Denied Domains

None - all network requests were properly authorized.

Firewall Activity by Workflow

• Changeset Generator: 4 requests (all allowed)

The firewall is functioning correctly with appropriate domain allowlisting. No suspicious or unauthorized network access attempts detected.

Tool Usage Statistics

Most frequently used tools across all workflows:

1. github - 23 calls (1 workflow)
2. TodoWrite - Multiple workflows
3. safeoutputs - 2 calls (1 workflow)
4. ast-grep - 1 call (1 workflow)

Affected Workflows

✅ Healthy Workflows (95% success or better)

• Tidy (11 runs, minimal errors)
• Smoke Claude (3 runs)
• Smoke Codex (3 runs)
• Changeset Generator
• Most scheduled workflows

⚠️ Workflows Needing Attention

• Documentation Unbloat: 1 failure - investigate scheduled job
• Go Pattern Detector: High error count but all runs successful (logging issue)
• Safe Output Health Monitor: High error-to-run ratio (logging artifact)

Recommendations

High Priority

1. Investigate Documentation Unbloat Failure

   • Review logs for §19446157378
   • Determine root cause of 37-second failure
   • Implement fix or adjust schedule if needed

2. Review Error Logging Practices

   • Many workflows log JSON messages at error severity level
   • Consider adjusting log levels for informational messages
   • Differentiate between actual errors and status messages

Medium Priority

3. Address EventEmitter Warnings

   • Node.js EventEmitter memory warnings in several workflows
   • Add emitter.setMaxListeners() calls where needed
   • Review resize listener usage patterns

4. Monitor Token Usage Trends

   • Daily token usage varies significantly (4M-9M tokens)
   • Monitor for cost efficiency opportunities
   • Consider implementing token budgets for high-usage workflows

Low Priority

5. Enhance Firewall Monitoring
   • Current firewall activity is minimal and appropriate
   • Continue monitoring for unauthorized access attempts
   • Consider expanding firewall logging for audit trails

Historical Context

This is the second audit in the index. Comparing with previous audits:

• Success Rate: Maintained at ~90% (excellent)
• Error Count: 556 errors (mostly logging artifacts)
• Cost Efficiency: $0.19 per run average (reasonable)

Next Steps

• Complete comprehensive audit of last 24 hours
• Generate health and cost trend charts
• Store analysis in cache memory
• Investigate Documentation Unbloat failure
• Review error logging practices across workflows
• Monitor next 24-hour period for patterns

---

References:

• §19447230606
• §19442042646
• §19446157378

AI generated by Agentic Workflow Audit Agent

[github-actions[bot]]

This discussion was automatically closed because it was created by an agentic workflow more than 1 week ago.