[Code Quality] Restore gh-aw MCP logs tool access for audit agent

[github-actions]

Description

The Agentic Workflow Audit Agent cannot access the gh-aw MCP server logs tool, preventing comprehensive workflow analysis including token usage, cost tracking, and error pattern identification.

Current Impact

From the 2026-01-27 Workflow Audit Report:

Cannot analyze:

• ❌ Detailed workflow execution logs
• ❌ Token usage statistics
• ❌ Actual costs
• ❌ Missing tool requests
• ❌ Specific error patterns from logs

Audit quality: Significantly degraded - relying only on GitHub API metadata and historical cache data

Root Cause

The gh-aw MCP server tools (agentic_workflows/status and agentic_workflows/logs) are not accessible during audit execution, either due to:

1. MCP server not configured for audit workflow
2. Permission/authentication issues
3. Tool registration not working

Suggested Changes

1. Verify MCP Configuration

Check audit workflow frontmatter (.github/workflows/audit-workflows.md):

tools:
  gh-aw:
    mode: remote  # or local
    toolsets: [logs, status]  # Ensure logs toolset is enabled

2. Add Fallback Error Handling

If MCP tools fail, the audit should:

• Log clear error message about missing tools
• Continue with available data sources
• Report degraded audit quality in summary

3. Validate MCP Server Health

Add a health check at audit start:

# Pseudo-code
if ! check_mcp_tool_available("agentic_workflows/logs"); then
    log_warning("MCP logs tool unavailable - audit will be incomplete")
fi

Files Affected

• .github/workflows/audit-workflows.md - audit workflow configuration
• pkg/cli/workflows/audit-*.md - if audit uses local workflow
• MCP server configuration files

Success Criteria

• Audit agent can successfully call agentic_workflows/logs tool
• Token usage and cost data available in audit reports
• Error patterns extracted from workflow logs
• Audit quality returns to "Complete" status
• Test with next scheduled audit run

Testing

# Manual test
gh aw run audit-workflows --engine copilot

# Verify logs access in output
# Should see token usage, cost data, and error patterns

Impact

High - Without this, audit reports lack critical data:

• Cannot track token usage trends
• Cannot calculate costs
• Cannot identify recurring error patterns
• Cannot provide actionable recommendations based on logs

Historical average success rate is only 12.7% - log access is critical for root cause analysis.

Source

Extracted from Agentic Workflow Audit Report discussion #11992 (2026-01-27).

Issue listed as #1 immediate action: "Restore MCP Logs Access".

Priority

High - Critical for comprehensive workflow health monitoring and debugging

AI generated by Discussion Task Miner - Code Quality Improvement Agent

• expires on Feb 17, 2026, 5:23 AM UTC