🔍 Agentic Workflow Audit Report - November 24, 2025

[github-actions[bot]]

🔍 Agentic Workflow Audit Report - November 24, 2025

Audit Period: Last 24 hours (2025-11-23 15:57 UTC to 2025-11-24 01:01 UTC)

Audit Summary

The audit system successfully analyzed workflow activity across the repository using the gh-aw MCP server. This report provides insights into workflow health, resource usage, errors, and recommendations for improvement.

📊 Key Metrics at a Glance

• Total Runs Analyzed: 35
• Active Workflows: 20+ distinct workflows
• Success Rate: 97.1% (34 successful, 1 failed)
• Total Runtime: 2.4 hours
• Total Token Usage: 7,777,415 tokens
• Total Cost: $6.47
• Average Turns per Run: 6.4
• Total Errors Logged: 359
• Total Warnings: 215
• Missing Tool Requests: 3

🎯 Overall Health Status

✅ HEALTHY - The agentic workflow system is operating normally with a 97.1% success rate. Most issues are minor and related to expected error handling patterns in validation scripts.

Breakdown by Engine

Engine	Runs	Success Rate	Avg Cost
Claude	17	100%	$0.36
Copilot	11	100%	N/A
Codex	5	100%	N/A
Unknown	2	50%	N/A

⚠️ Issues Found

1. Missing Tools & MCP Failures

Missing Tools Summary

Tool Name	Request Count	Workflows Affected	Root Cause
playwright	2	Smoke Copilot	MCP timeout: Request timed out after initialization
serena	3	Smoke Copilot	Missing dependency: spawn uvx ENOENT

Impact: Low - Only affects smoke test workflows, not production workflows.

Recommendations:

1. Playwright Timeout: Increase MCP server timeout for playwright initialization or investigate slow startup
2. Serena MCP: Install uvx dependency or remove serena MCP configuration if not needed

2. Error Analysis

Error Breakdown

Critical Errors: 0

Pattern-Based Errors: Most errors (89 instances) are from the Security Fix PR workflow's validation script scanning for error patterns in logs. These are expected and part of the validation process, not actual failures.

Notable Real Errors:

1. TOML Parse Errors (14 occurrences) - Smoke Codex workflow

   • Pattern: TOML parse error at line 30, column 109
   • Impact: Low - appears to be a warning/detection pattern, workflow completes successfully

2. MCP Timeouts (4 occurrences) - Smoke Copilot

   • Pattern: MCP error -32001: Request timed out
   • Impact: Medium - blocks playwright MCP from starting

3. Spawn uvx ENOENT (3 occurrences) - Smoke Copilot

   • Pattern: Error: spawn uvx ENOENT
   • Impact: Medium - prevents serena MCP from starting

4. JSON Parse Errors (6 occurrences) - Q workflow

   • Pattern: Unexpected non-whitespace character after JSON
   • Impact: Low - workflow completes successfully

3. Failed Workflows

Failed Run Details

1 Failure Detected:

• Workflow: Claude Command Processor - /cloclo
• Run ID: §19614059145
• Event: pull_request_review_comment
• Branch: refactor/deduplicate-participant-config-parsing-443b1b13899c538b
• Duration: 2.4m
• Error Count: 1

Recommendation: Review the specific failure logs for run §19614059145 to determine root cause.

💰 Cost Analysis

Top 5 Most Expensive Workflows (Last 24h)

Workflow	Cost	Tokens	Turns	Duration
Documentation Unbloat	$1.33	2.3M	42	8.8m
Claude Command Processor - /cloclo	$0.83	1.4M	26	7.3m
Copilot Agent PR Analysis	$0.60	581K	14	3.5m
Smoke Claude	$0.50	599K	13	4.3m
Smoke Claude (Run 2)	$0.44	399K	15	3.5m

Total 24h Cost: $6.47 across 35 runs
Daily Projection: ~$6.50/day
Monthly Projection: ~$195/month (at current activity level)

Cost Optimization Opportunities

1. Documentation Unbloat (42 turns, $1.33) - High turn count suggests potential for optimization
2. Smoke Tests - Multiple smoke test runs contribute ~$1.50/day. Consider reducing frequency if appropriate.

🔧 Tool Usage Statistics

Top Tools by Call Count

Tool	Total Calls	Used in Runs	Max Output Size
github	277	12	N/A
TodoWrite	62	12	2.2KB
safeoutputs	45	12	N/A
Read	34	13	16KB
Edit	16	2	1.7KB
github_search_pull_requests	13	9	2.2KB
playwright_browser_navigate	9	9	2.2KB

📈 Workflow Activity Patterns

Most Active Workflows (Last 24h)

1. Smoke Tests (Claude, Copilot, Codex) - 9 runs
2. Tidy - 4 runs
3. Go Pattern Detector - 4 runs
4. Security Fix PR - 3 runs
5. Changeset Generator - 3 runs

Event Triggers

• Schedule: 18 runs (51%)
• Push: 10 runs (29%)
• Pull Request: 5 runs (14%)
• Other: 2 runs (6%)

🎯 Recommendations

High Priority

1. Fix Serena MCP Dependency: Install uvx or remove serena MCP configuration to eliminate spawn errors
2. Investigate Playwright Timeout: Increase timeout or optimize playwright MCP server startup
3. Review Failed Run: Investigate §19614059145 for root cause analysis

Medium Priority

4. Optimize High-Turn Workflows: Review Documentation Unbloat (42 turns) for potential consolidation
5. Monitor Token Usage: Documentation Unbloat used 2.3M tokens in one run - consider if this is expected
6. TOML Parse Warnings: Investigate repeated TOML parse errors in Smoke Codex

Low Priority

7. Error Message Cleanup: Many "errors" are false positives from validation scripts - consider refining error detection patterns
8. JSON Parse Handling: Q workflow has JSON parse warnings but completes successfully - may need better error handling

📊 Historical Context

This is the initial audit report. Future audits will include:

• Week-over-week comparison
• Trend analysis for success rates
• Cost trend charts
• Error pattern evolution

✅ Next Steps

• Install missing dependencies for serena MCP
• Investigate playwright MCP timeout issues
• Review and analyze failed run §19614059145
• Monitor Documentation Unbloat workflow for optimization opportunities
• Schedule next audit for 2025-11-25

---

Audit Completed: 2025-11-24 01:03 UTC
Audit Tool: gh-aw MCP Server v1.0
Data Source: Workflow logs from /tmp/gh-aw/aw-mcp/logs

---

References:

• §19614059145 - Failed run
• §19617967406 - Highest cost run
• §19619460418 - MCP timeout example

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.