[portfolio] Portfolio Dashboard - 2026-03-09

[github-actions[bot]]

This week's analysis covers 287 workflow runs across 78 unique workflows over a ~2-day window (2026-03-07 to 2026-03-09), with $45.62 total cost measured from 24 runs that have full token/cost instrumentation. The headline finding: 48% of all runs failed — a significant operational health issue concentrated in a handful of always-failing workflows.

⚠️ Data Note: Cost data (estimated_cost) is only populated for 24 of 287 runs. Many workflows run without cost tracking (likely non-LLM or lightweight jobs). The $45.62 figure covers only instrumented runs. The analysis period is ~2 days, not 30 days.

---

Visual Summary

Cost Trends

Key Insights:

• Period covered: 2026-03-07 → 2026-03-09 (2-day snapshot)
• Total tracked cost: $45.62 across 24 instrumented runs
• Peak day: 2026-03-08 (~$36) — normal scheduled batch

Top Cost Drivers

Top 3 workflows account for ~42% of tracked cost:

1. go-logger-enhancement.md — $8.16/period (2 runs, 0% failure) ✅
2. copilot-session-insights.md — $5.29/period (2 runs, 0% failure) ✅
3. sergo-serena-go-expert.md — $4.01/period (2 runs, 0% failure) ✅

Failure Analysis

48% of all runs failed — driven by a small set of 100%-failing workflows:

• 12 workflows with ≥30% failure rate
• 10 workflows at exactly 100% failure (zero successful runs ever)
• Wasted infrastructure cost minimal (most failing workflows lack cost instrumentation), but wasted CI minutes are significant

Overall Health

• ✅ Success: 50.2% (144 runs)
• ❌ Failure: 48.1% (138 runs)
• ⏸️ Cancelled: 1.4% (4 runs)
• 🔄 Pending: 0.3% (1 run)

---

💰 Cost Reduction Opportunities

Priority: Fix/Disable 100%-Failing Workflows to recover CI minutes and improve signal quality

Since most failing workflows have no cost instrumentation, direct dollar savings are limited, but recovering CI capacity and reducing noise is high value.

Strategy 1: Disable 100%-Failing Workflows (Critical)

These workflows have never succeeded in the observed period and should be investigated and either fixed or disabled:

Workflow	Runs	Failures	Failure Rate
Issue Monster	65	65	100%
AI Moderator	22	22	100%
Smoke Codex	14	14	100%
PR Triage Agent	7	7	100%
Smoke Update Cross-Repo PR	3	3	100%
The Great Escapi	2	2	100%
Security Alert Burndown	2	2	100%
Duplicate Code Detector	2	2	100%
Daily Issues Report Generator	2	2	100%
Organization Health Report	1	1	100%

Action: Audit each workflow. For event-driven workflows (Issue Monster, AI Moderator), check trigger configuration. For smoke tests (Smoke Codex), check if the engine/environment is broken.

Strategy 2: Investigate High-Cost Workflows for Frequency Tuning

These workflows have full cost data and run on schedule. All currently healthy (0% failure), but worth reviewing frequency vs. value:

Workflow	Cost/2-days	Runs	Est. Monthly
Go Logger Enhancement	$8.16	2	~$122
Copilot Session Insights	$5.29	2	~$79
Sergo - Serena Go Expert	$4.01	2	~$60
Daily Documentation Updater	$2.62	2	~$39
Daily Security Red Team Agent	$2.34	1	~$35
Documentation Unbloat	$2.24	2	~$34
Daily Code Metrics Agent	$2.04	1 (100% fail!)	~$31 wasted

Highest priority: Daily Code Metrics and Trend Tracking Agent — $2.04 cost with 100% failure rate. Fix or disable immediately.

Frequency review: Workflows running daily that cost >$2/run could save 50-75% by switching to weekly. Example: if Go Logger Enhancement runs daily at ~$4/run → switching to weekly saves ~$84/month.

Strategy 3: Investigate High-Volume Event Workflows

Issue Monster (65 runs) and AI Moderator (22 runs) are event-driven and both failing 100%. These are high-volume and completely broken.

Issue Monster — 65 failures is the single largest contributor to the 48% failure rate. Fixing this one workflow alone would drop the portfolio failure rate from 48% to ~26%.

AI Moderator — 22 failures, likely similar root cause.

Recommended action: Check error logs for these two workflows first.

---

🎯 Priority Actions

1. CRITICAL — Fix or disable Issue Monster: 65 failed runs is the largest single source of failures; fixing it alone cuts overall failure rate nearly in half. Check .github/workflows/issue-monster.md for trigger and permission issues.

2. CRITICAL — Fix or disable AI Moderator: 22 failed runs, likely same root cause as Issue Monster (event-driven workflows with misconfigured triggers or auth).

3. HIGH — Disable Daily Code Metrics and Trend Tracking Agent: $2.04/run with 100% failure rate = pure waste. Either fix the underlying issue or disable until repaired.

4. HIGH — Audit Smoke Codex (14 failures, 100%): If Codex engine is broken/unavailable, all Codex smoke tests will continue failing indefinitely.

5. MEDIUM — Review frequency of Go Logger Enhancement ($8.16/2 days): If this runs daily at that cost, monthly spend is ~$122 for a single workflow. Verify business need for daily cadence.

---

📈 Data Quality

• Period Analyzed: 2026-03-07 to 2026-03-09 (~2 days)
• Total Runs: 287 workflow runs across 78 unique workflows
• Cost Coverage: 24/287 runs have full cost data ($45.62 total)
• Confidence: Medium — 2-day snapshot provides clear failure patterns but cost projections are extrapolated
• Limitation: ~90% of runs lack cost instrumentation; monthly cost estimates should be treated as approximations

---

Methodology: Analysis based on pre-downloaded workflow execution data from gh aw logs. Costs calculated from actual token usage for instrumented runs. Failure rates computed from all 287 runs regardless of cost instrumentation.

References:

• §22845995633 — This analysis run

AI generated by Automated Portfolio Analyst · history

• expires on Mar 10, 2026, 9:26 AM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-03-10T09:26:20.499Z.

Closed by Workflow