Workflow Skill Extractor Report — 2026-03-22

[github-actions[bot]]

🎯 Executive Summary

This analysis reviewed all 177 agentic workflow files in .github/workflows/ against 62 existing shared components. Three high-value opportunities for new shared components were identified and corresponding actionable issues were created. The repository already has a strong shared-component culture — shared/reporting.md is imported by 97 of 177 workflows — but several clear duplication patterns remain unaddressed.

The three recommendations focus on: eliminating verbatim code duplication in log-download steps, providing a unified MCP usage guide for the widely-used agentic-workflows tool, and consolidating the recurring PR code review configuration pattern.

Key Statistics:

• Total workflows analyzed: 177
• Existing shared components: 62
• New skills identified: 8 (3 high-priority, 2 medium, 3 low)
• Estimated total lines saved (top 3): ~250–350 lines (direct); 500+ lines if fully adopted

---

📊 Analysis Overview

Most-Used Shared Components

Shared Component	Workflows
shared/reporting.md	97
shared/mcp/serena-go.md	17
shared/jqschema.md	18
shared/python-dataviz.md	8
shared/gh.md	6
shared/trends.md	5
shared/github-queries-mcp-script.md	5
shared/copilot-pr-analysis-base.md	4

Workflows With No Imports

56 workflows do not import any shared component. Most are event-driven (slash commands, issue events) rather than scheduled, and generally too specific to benefit from additional shared components beyond shared/reporting.md.

Key Tool & Feature Usage

Pattern	Workflow Count
agentic-workflows: tool	28
copilot-requests: true feature	41
upload-asset: safe-output	22
close-older-discussions: true	55
close-older-issues: true	17
cache-memory: tool	57
repo-memory: tool	26
messages: run-started:	41
submit-pull-request-review:	7

---

🔍 Identified Skills

High Priority Skills

1. Agentic Workflow Logs Pre-fetch (24h)

• Frequency: 2 exact duplicates + 4 near-identical
• Size: ~15 lines per workflow
• Priority: 🔴 High
• Description: Identical tools: block + steps: entry to pre-download last 24h agentic workflow logs before agent execution
• Workflows: safe-output-health, daily-safe-output-optimizer, portfolio-analyst, daily-copilot-token-report, daily-integrity-analysis, prompt-clustering-analysis
• Recommendation: Extract to shared/aw-logs-24h-fetch.md

2. Agentic Workflows MCP Usage Guide

• Frequency: 28 workflows use the tool; 8 copy-paste the safety warning verbatim
• Size: ~20 lines of guidance per workflow
• Priority: 🔴 High
• Description: Tool configuration + "DO NOT USE GH AW DIRECTLY" warning + status/logs/audit tool usage instructions — analogous to shared/mcp/serena-go.md
• Workflows: All 28 agentic-workflows: consumers
• Recommendation: Extract to shared/agentic-workflows-mcp.md

3. PR Code Review Base Configuration

• Frequency: 7 workflows (4 production + 3 smoke tests)
• Size: ~8–10 lines per workflow
• Priority: 🔴 High
• Description: Consistent cache-memory + GitHub PR tools + create-pull-request-review-comment (side: RIGHT) + submit-pull-request-review configuration
• Workflows: grumpy-reviewer, pr-nitpick-reviewer, security-review, daily-safe-output-integrator, smoke-copilot, smoke-copilot-arm, smoke-claude
• Recommendation: Extract to shared/pr-code-review-config.md

Medium Priority Skills

4. Meta-Orchestrator Repo-Memory Config

• Frequency: 3 workflows
• Size: ~5–8 lines
• Description: repo-memory: branch-name: memory/meta-orchestrators shared by agent-performance-analyzer, workflow-health-manager, metrics-collector
• Recommendation: Could be added to shared/meta-orchestrators.md

5. Rotating Strategy Cache Pattern

• Frequency: 3 workflows
• Size: ~30–50 lines of strategy-selection logic
• Description: Pattern for rotating between proven and new analysis strategies using cache-memory (70%/30% split); appears in schema-consistency-checker, sergo, slide-deck-maintainer
• Recommendation: Could be extracted as shared/rotating-cache-strategy.md

Low Priority Skills

• make build + setup-go binary build steps — 5 workflows with identical pattern (others vary); would save ~12 lines × 5 = 60 lines
• Daily discussion with close-older-discussions: true — 55 workflows, but each has a unique title-prefix making clean extraction challenging
• "Current Context" prompt block — ~40 workflows have the ## Current Context / $\{\{ github.repository }} / $\{\{ github.run_id }} / date pattern, but it's only 4–5 lines and context-specific

---

💡 Detailed Recommendations

Recommendation 1: Agentic Workflow Logs Pre-fetch

Full Details

Current State (identical in 2 workflows):

# safe-output-health.md AND daily-safe-output-optimizer.md — IDENTICAL
tools:
  agentic-workflows:
  cache-memory: true
  timeout: 300

steps:
  - name: Download logs from last 24 hours
    env:
      GH_TOKEN: $\{\{ secrets.GITHUB_TOKEN }}
    run: ./gh-aw logs --start-date -1d -o /tmp/gh-aw/aw-mcp/logs

Proposed shared/aw-logs-24h-fetch.md:

---
tools:
  agentic-workflows:
  cache-memory: true
  timeout: 300

steps:
  - name: Download logs from last 24 hours
    env:
      GH_TOKEN: $\{\{ secrets.GITHUB_TOKEN }}
    run: ./gh-aw logs --start-date -1d -o /tmp/gh-aw/aw-mcp/logs
---

## Agentic Workflow Logs (Last 24h)

Logs are pre-downloaded to `/tmp/gh-aw/aw-mcp/logs/`. Use MCP server tools for additional queries — do NOT run `./gh-aw` directly.

After migration:

# safe-output-health.md — simplified
imports:
  - shared/aw-logs-24h-fetch.md
  - shared/jqschema.md
  - shared/reporting.md
```

**Impact:** 15 lines removed from 2 exact matches; template for 4+ others.

</details>

#### Recommendation 2: Agentic Workflows MCP Usage Guide

<details>
<summary><b>Full Details</b></summary>

**Current State (8 variants of same warning):**

```
# safe-output-health.md, daily-safe-output-optimizer.md:
- DO NOT ATTEMPT TO USE GH AW DIRECTLY, it is not authenticated. Use the MCP server instead.

# q.md:
**DO NOT ATTEMPT TO USE GH AW DIRECTLY** - it is not authenticated. Use the MCP server instead.

# daily-rendering-scripts-verifier.md:
DO NOT USE `gh aw` CLI directly for GitHub API operations — it is not authenticated.

Proposed shared/agentic-workflows-mcp.md (modeled after shared/mcp/serena-go.md):

---
tools:
  agentic-workflows:
---

## Agentic Workflows MCP Server

⚠️ Do NOT use `./gh-aw` or `gh aw` CLI commands — NOT authenticated in agent environment. Use MCP tools.

### Tools: `status`, `logs`, `audit`

Migration: Remove agentic-workflows: from tools: and the copy-pasted warning from each workflow body; add imports: - shared/agentic-workflows-mcp.md.

Impact: Centralizes safety warning; consistent tool documentation for all 28 consumers.

Recommendation 3: PR Code Review Base Configuration

Full Details

Common pattern across 4 production review workflows:

tools:
  cache-memory: true
  github:
    toolsets: [pull_requests, repos]
safe-outputs:
  create-pull-request-review-comment:
    max: 5-10   # varies
    side: "RIGHT"
  submit-pull-request-review:
    max: 1

Proposed shared/pr-code-review-config.md:

---
tools:
  cache-memory: true
  github:
    toolsets: [pull_requests, repos]
safe-outputs:
  create-pull-request-review-comment:
    side: "RIGHT"
  submit-pull-request-review:
    max: 1
---

## PR Code Review Configuration

Use `cache-memory` to persist review history at `/tmp/gh-aw/cache-memory/pr-{number}.json`. Batch all comments before calling `submit-pull-request-review` once.

Impact: Consistent side: "RIGHT" enforcement; reduces boilerplate for new review agents.

---

📈 Impact Analysis

By Category

Category	Skills Found	Estimated Lines Saved
Tool Configurations	2	~200 lines
Setup Steps	1	~90 lines
Config Consolidation	1	~56 lines
Prompt/Strategy Patterns	2	~100 lines
Minor / Low Priority	3	~80 lines

By Priority

Priority	Skills	Lines Saved	Workflows Affected
High 🔴	3	~250–350	37
Medium 🟡	2	~100	6
Low 🟢	3	~80	12

---

✅ Created Issues

This analysis has created the following actionable issues:

1. [refactoring] Extract Agentic Workflow Logs Pre-fetch into shared component #22286 — Extract Agentic Workflow Logs Pre-fetch into shared component
2. [refactoring] Extract Agentic Workflows MCP usage guide into shared component #22287 — Extract Agentic Workflows MCP usage guide into shared component
3. [refactoring] Extract PR Code Review base configuration into shared component #22288 — Extract PR Code Review base configuration into shared component

---

🎯 Next Steps

1. Review and prioritize the 3 created issues
2. Start with [refactoring] Extract Agentic Workflow Logs Pre-fetch into shared component #22286 (clearest ROI, 2 exact duplicates to eliminate immediately)
3. Then tackle [refactoring] Extract Agentic Workflows MCP usage guide into shared component #22287 (highest adoption potential, 28 workflows)
4. Then [refactoring] Extract PR Code Review base configuration into shared component #22288 (smaller but improves new-reviewer onboarding)
5. Monitor for new skills in future workflow additions (especially as agentic-workflows tool adoption grows)
6. Consider scheduling next extractor run in ~1 month

---

📚 Methodology

• Analyzed 177 agentic workflow markdown files
• Reviewed 62 existing shared components to avoid redundancy
• Identified skill patterns across 4 categories: tool configs, prompt instructions, setup steps, data processing
• Prioritized based on frequency (workflows affected), duplication clarity (exact vs. near-identical), and complexity of extraction
• Generated top 3 actionable recommendations with implementation plans

Analysis Date: 2026-03-22
Analyzer: Workflow Skill Extractor v1.0

References:

• Workflow Skill Extractor run §23407097688

AI generated by Workflow Skill Extractor · history

• expires on Mar 29, 2026, 4:26 PM UTC

[github-actions[bot]]

This discussion has been marked as outdated by Workflow Skill Extractor.

A newer discussion is available at Discussion #23103.