CI/CD Pipeline Monitor

I'll analyze your CI/CD pipeline metrics, track build success rates, identify flaky tests, and provide performance trend analysis.

Arguments: $ARGUMENTS - pipeline platform (github, gitlab, circle), time range, or specific build numbers

Monitoring Philosophy

Data-Driven Insights: Identify trends, not just failures
Flaky Test Detection: Find tests that fail inconsistently
Performance Tracking: Monitor build duration over time
Success Rate Metrics: Track reliability trends
Multi-Platform: Support GitHub Actions, GitLab CI, CircleCI, Jenkins

Token Optimization

This skill uses efficient patterns to minimize token consumption during CI/CD pipeline monitoring and analysis.

Optimization Strategies

1. CI Platform Detection Caching (Saves 600 tokens per invocation)

Cache detected CI platform and configuration paths:

CACHE_FILE=".claude/cache/pipeline-monitor/platform.json"
CACHE_TTL=86400  # 24 hours (CI config rarely changes)

mkdir -p .claude/cache/pipeline-monitor

if [ -f "$CACHE_FILE" ]; then
    CACHE_AGE=$(($(date +%s) - $(stat -c %Y "$CACHE_FILE" 2>/dev/null || stat -f %m "$CACHE_FILE" 2>/dev/null)))

    if [ $CACHE_AGE -lt $CACHE_TTL ]; then
        # Use cached platform info
        CI_PLATFORM=$(jq -r '.platform' "$CACHE_FILE")
        CI_CONFIG=$(jq -r '.config_file' "$CACHE_FILE")
        API_ENDPOINT=$(jq -r '.api_endpoint' "$CACHE_FILE")

        echo "Using cached CI platform: $CI_PLATFORM"
        SKIP_DETECTION="true"
    fi
fi

# First run: detect and cache
if [ "$SKIP_DETECTION" != "true" ]; then
    detect_ci_platform  # Check for .github/workflows, .gitlab-ci.yml, etc.

    # Cache results
    jq -n \
        --arg platform "$CI_PLATFORM" \
        --arg config "$CI_CONFIG" \
        --arg api "$API_ENDPOINT" \
        '{platform: $platform, config_file: $config, api_endpoint: $api}' \
        > "$CACHE_FILE"
fi

Savings: 600 tokens (no repeated directory scans, no file existence checks)

2. API Response Caching (Saves 80%)

Cache CI/CD API responses to avoid repeated network calls:

# Cache API responses (5 minute TTL for build data)
API_CACHE=".claude/cache/pipeline-monitor/builds-cache.json"
CACHE_TTL=300  # 5 minutes (builds change frequently)

if [ -f "$API_CACHE" ]; then
    CACHE_AGE=$(($(date +%s) - $(stat -c %Y "$API_CACHE" 2>/dev/null || stat -f %m "$API_CACHE" 2>/dev/null)))

    if [ $CACHE_AGE -lt $CACHE_TTL ]; then
        echo "Using cached build data ($(($CACHE_AGE / 60)) minutes old)"
        cat "$API_CACHE"
        exit 0
    fi
fi

# Fetch and cache
case "$CI_PLATFORM" in
    github-actions)
        gh api repos/:owner/:repo/actions/runs --jq '.workflow_runs[:50]' > "$API_CACHE"
        ;;
    gitlab-ci)
        curl -H "PRIVATE-TOKEN: $GITLAB_TOKEN" \
            "$GITLAB_API_URL/projects/$PROJECT_ID/pipelines?per_page=50" > "$API_CACHE"
        ;;
esac

Savings: 80% when cache valid (no API calls, instant response: 2,000 → 400 tokens)

3. Sample-Based Metrics Analysis (Saves 75%)

Analyze last 50 builds, not entire history:

# Efficient: Only analyze recent builds
ANALYSIS_LIMIT="${ANALYSIS_LIMIT:-50}"  # Default: last 50 builds

analyze_build_metrics() {
    local builds_json="$1"

    # Extract key metrics only (not full build data)
    TOTAL_BUILDS=$(jq 'length' "$builds_json")
    SUCCESS_COUNT=$(jq '[.[] | select(.conclusion == "success")] | length' "$builds_json")
    FAILURE_COUNT=$(jq '[.[] | select(.conclusion == "failure")] | length' "$builds_json")
    SUCCESS_RATE=$(echo "scale=2; $SUCCESS_COUNT * 100 / $TOTAL_BUILDS" | bc)

    # Average duration (sample-based)
    AVG_DURATION=$(jq '[.[] | .run_duration_ms] | add / length / 1000' "$builds_json")

    echo "Build Metrics (last $ANALYSIS_LIMIT builds):"
    echo "  Success Rate: ${SUCCESS_RATE}%"
    echo "  Total: $TOTAL_BUILDS | Success: $SUCCESS_COUNT | Failures: $FAILURE_COUNT"
    echo "  Avg Duration: ${AVG_DURATION}s"

    echo ""
    echo "Use --all-history for complete analysis"
}

Savings: 75% (analyze 50 vs 500+ builds: 3,000 → 750 tokens)

4. Flaky Test Pattern Detection (Saves 85%)

Use statistical sampling to identify flaky tests:

# Efficient: Pattern-based flaky detection (not exhaustive analysis)
detect_flaky_tests() {
    local builds_json="$1"

    echo "Detecting flaky tests..."

    # Extract failed test names from recent failures
    FAILED_TESTS=$(jq -r '.[] |
        select(.conclusion == "failure") |
        .jobs[].steps[] |
        select(.conclusion == "failure") |
        .name' "$builds_json" | sort | uniq -c | sort -rn)

    # Identify tests that failed 2-4 times (flaky pattern)
    FLAKY_CANDIDATES=$(echo "$FAILED_TESTS" | awk '$1 >= 2 && $1 <= 4')

    if [ -n "$FLAKY_CANDIDATES" ]; then
        echo "Potential flaky tests (failed 2-4 times):"
        echo "$FLAKY_CANDIDATES" | head -10 | while read count name; do
            PCT=$(echo "scale=0; $count * 100 / $TOTAL_BUILDS" | bc)
            echo "  - $name (${PCT}% failure rate)"
        done
    else
        echo "✓ No flaky tests detected"
    fi

    echo ""
    echo "Use --detailed-flaky for statistical analysis"
}

Savings: 85% (pattern detection vs full statistical analysis: 2,000 → 300 tokens)

5. Bash-Based Log Parsing (Saves 70%)

Parse CI logs with grep/awk instead of full reads:

# Efficient: Grep for error patterns (not full log analysis)
analyze_failure_patterns() {
    local log_file="$1"

    echo "Analyzing failure patterns..."

    # Count error types (efficient grep)
    TIMEOUT_ERRORS=$(grep -c "timeout\|ETIMEDOUT" "$log_file" 2>/dev/null || echo "0")
    OOM_ERRORS=$(grep -c "out of memory\|OOM" "$log_file" 2>/dev/null || echo "0")
    NETWORK_ERRORS=$(grep -c "ECONNREFUSED\|network" "$log_file" 2>/dev/null || echo "0")
    TEST_FAILURES=$(grep -c "FAILED\|AssertionError" "$log_file" 2>/dev/null || echo "0")

    # Summary (no full log output)
    echo "Error Distribution:"
    [ $TIMEOUT_ERRORS -gt 0 ] && echo "  - Timeouts: $TIMEOUT_ERRORS"
    [ $OOM_ERRORS -gt 0 ] && echo "  - Out of Memory: $OOM_ERRORS"
    [ $NETWORK_ERRORS -gt 0 ] && echo "  - Network: $NETWORK_ERRORS"
    [ $TEST_FAILURES -gt 0 ] && echo "  - Test Failures: $TEST_FAILURES"
}

Savings: 70% vs full log parsing (grep counts vs full read: 1,500 → 450 tokens)

6. Progressive Metrics Reporting (Saves 60%)

Default to summary, provide detailed analysis on demand:

DETAIL_LEVEL="${DETAIL_LEVEL:-summary}"

case "$DETAIL_LEVEL" in
    summary)
        # Quick metrics (400 tokens)
        echo "Success Rate: ${SUCCESS_RATE}%"
        echo "Last Build: $(jq -r '.[0].conclusion' builds.json)"
        echo "Flaky Tests: $FLAKY_COUNT"
        echo ""
        echo "Use --detailed for complete analysis"
        ;;

    detailed)
        # Medium detail (1,200 tokens)
        show_build_metrics
        show_flaky_tests
        show_duration_trend
        ;;

    full)
        # Complete analysis (2,500 tokens)
        show_all_builds
        show_detailed_flaky_analysis
        show_failure_patterns
        show_recommendations
        ;;
esac

Savings: 60% for default runs (400 vs 1,200-2,500 tokens)

7. GitHub CLI Integration (Saves 75%)

Use gh CLI instead of REST API for GitHub Actions:

# Efficient: gh CLI with JSON output (no auth setup needed)
if [ "$CI_PLATFORM" = "github-actions" ]; then
    # Single command, JSON output
    gh run list --limit 50 --json conclusion,status,name,startedAt,durationMs \
        > "$API_CACHE"

    # Parse directly (no intermediate processing)
    SUCCESS_RATE=$(jq '[.[] | select(.conclusion == "success")] | length / length * 100' "$API_CACHE")

    echo "GitHub Actions: $SUCCESS_RATE% success rate (last 50 runs)"
fi

Savings: 75% vs manual REST API calls (gh CLI handles auth, pagination: 1,200 → 300 tokens)

Cache Invalidation

Caches are invalidated when:

CI configuration files modified
5 minutes elapsed (time-based for build data)
24 hours elapsed (time-based for platform detection)
User runs --clear-cache or --fresh flag
New builds detected

Real-World Token Usage

Typical monitoring workflow:

Quick status check: 400-800 tokens
- Cached platform: 100 tokens
- Cached build data (< 5 min): 200 tokens
- Success rate calculation: 150 tokens
- Summary output: 200 tokens
First-time analysis: 1,200-1,800 tokens
- Platform detection: 300 tokens
- API fetch (50 builds): 400 tokens
- Metrics calculation: 300 tokens
- Flaky test detection: 400 tokens
- Summary: 200 tokens
Detailed analysis: 1,800-2,500 tokens
- All basic metrics: 800 tokens
- Detailed flaky analysis: 600 tokens
- Duration trends: 400 tokens
- Recommendations: 300 tokens
Full historical analysis: 2,500-3,500 tokens
- Only when explicitly requested with --full flag

Average usage distribution:

60% of runs: Cached quick check (400-800 tokens) ✅ Most common
25% of runs: First-time analysis (1,200-1,800 tokens)
10% of runs: Detailed analysis (1,800-2,500 tokens)
5% of runs: Full historical (2,500-3,500 tokens)

Expected token range: 400-2,500 tokens (50% reduction from 800-5,000 baseline)

Progressive Disclosure

Three levels of monitoring:

Default (summary): Quick health check

claude "/pipeline-monitor"
# Shows: success rate, last build status, flaky count
# Tokens: 400-800

Detailed (trends): Performance analysis

claude "/pipeline-monitor --detailed"
# Shows: metrics, flaky tests, duration trends
# Tokens: 1,200-1,800

Full (historical): Complete pipeline analysis

claude "/pipeline-monitor --full"
# Shows: all builds, detailed flaky analysis, recommendations
# Tokens: 2,500-3,500

Implementation Notes

Key patterns applied:

✅ CI platform detection caching (600 token savings)
✅ API response caching (80% reduction when cached)
✅ Sample-based metrics (75% savings - last 50 builds)
✅ Flaky test pattern detection (85% savings)
✅ Bash-based log parsing (70% savings)
✅ Progressive metrics reporting (60% savings)
✅ GitHub CLI integration (75% savings for GitHub Actions)

Cache locations:

.claude/cache/pipeline-monitor/platform.json - CI platform and config (24 hour TTL)
.claude/cache/pipeline-monitor/builds-cache.json - Build data (5 minute TTL)
.claude/cache/pipeline-monitor/flaky-tests.json - Flaky test patterns (1 hour TTL)

Flags:

--detailed - Medium detail level (trends + flaky tests)
--full - Complete historical analysis
--fresh - Bypass all caches
--limit=<N> - Number of builds to analyze (default: 50)
--clear-cache - Force cache invalidation

Supported platforms:

GitHub Actions (gh CLI, GitHub API)
GitLab CI (GitLab API)
CircleCI (CircleCI API)
Jenkins (Jenkins API, log files)
Travis CI (Travis API)
Azure Pipelines (Azure DevOps API)

Phase 1: Pipeline Detection

First, I'll detect your CI/CD platform:

#!/bin/bash
# Detect CI/CD platform and configuration

detect_ci_platform() {
    echo "=== CI/CD Platform Detection ==="
    echo ""

    CI_PLATFORM=""
    CI_CONFIG=""

    # GitHub Actions
    if [ -d ".github/workflows" ]; then
        CI_PLATFORM="github-actions"
        CI_CONFIG=$(find .github/workflows -name "*.yml" -o -name "*.yaml" | head -1)
        echo "✓ Detected: GitHub Actions"
        echo "  Config: $CI_CONFIG"

    # GitLab CI
    elif [ -f ".gitlab-ci.yml" ]; then
        CI_PLATFORM="gitlab-ci"
        CI_CONFIG=".gitlab-ci.yml"
        echo "✓ Detected: GitLab CI"
        echo "  Config: $CI_CONFIG"

    # CircleCI
    elif [ -f ".circleci/config.yml" ]; then
        CI_PLATFORM="circleci"
        CI_CONFIG=".circleci/config.yml"
        echo "✓ Detected: CircleCI"
        echo "  Config: $CI_CONFIG"

    # Jenkins
    elif [ -f "Jenkinsfile" ]; then
        CI_PLATFORM="jenkins"
        CI_CONFIG="Jenkinsfile"
        echo "✓ Detected: Jenkins"
        echo "  Config: $CI_CONFIG"

    # Travis CI
    elif [ -f ".travis.yml" ]; then
        CI_PLATFORM="travis"
        CI_CONFIG=".travis.yml"
        echo "✓ Detected: Travis CI"
        echo "  Config: $CI_CONFIG"

    # Azure Pipelines
    elif [ -f "azure-pipelines.yml" ]; then
        CI_PLATFORM="azure-pipelines"
        CI_CONFIG="azure-pipelines.yml"
        echo "✓ Detected: Azure Pipelines"
        echo "  Config: $CI_CONFIG"

    else
        echo "⚠️  No CI/CD configuration detected"
        echo "Supported platforms: GitHub Actions, GitLab CI, CircleCI, Jenkins"
        exit 1
    fi

    echo ""
    echo "$CI_PLATFORM|$CI_CONFIG"
}

CI_INFO=$(detect_ci_platform)
CI_PLATFORM=$(echo "$CI_INFO" | cut -d'|' -f1)
CI_CONFIG=$(echo "$CI_INFO" | cut -d'|' -f2)

Phase 2: Build History Analysis

I'll fetch and analyze recent build history:

#!/bin/bash
# Fetch build history from CI platform

fetch_build_history() {
    local platform="$1"
    local limit="${2:-50}"

    echo "=== Fetching Build History ==="
    echo ""

    case "$platform" in
        github-actions)
            # Use GitHub CLI to fetch workflow runs
            if command -v gh &> /dev/null; then
                echo "Fetching last $limit GitHub Actions runs..."
                gh run list --limit "$limit" --json status,conclusion,name,createdAt,updatedAt,databaseId > /tmp/ci_builds.json

                # Parse and display summary
                TOTAL=$(jq length /tmp/ci_builds.json)
                SUCCESS=$(jq '[.[] | select(.conclusion=="success")] | length' /tmp/ci_builds.json)
                FAILURE=$(jq '[.[] | select(.conclusion=="failure")] | length' /tmp/ci_builds.json)
                SUCCESS_RATE=$(echo "scale=2; $SUCCESS * 100 / $TOTAL" | bc)

                echo "Total runs: $TOTAL"
                echo "Successful: $SUCCESS ($SUCCESS_RATE%)"
                echo "Failed: $FAILURE"
                echo "✓ Build history fetched"
            else
                echo "⚠️  gh CLI not installed. Install: https://cli.github.com/"
                exit 1
            fi
            ;;

        gitlab-ci)
            # Use GitLab CLI to fetch pipelines
            if command -v glab &> /dev/null; then
                echo "Fetching last $limit GitLab CI pipelines..."
                glab ci list --per-page "$limit" --output json > /tmp/ci_builds.json

                # Parse and display summary
                TOTAL=$(jq length /tmp/ci_builds.json)
                SUCCESS=$(jq '[.[] | select(.status=="success")] | length' /tmp/ci_builds.json)
                FAILURE=$(jq '[.[] | select(.status=="failed")] | length' /tmp/ci_builds.json)
                SUCCESS_RATE=$(echo "scale=2; $SUCCESS * 100 / $TOTAL" | bc)

                echo "Total pipelines: $TOTAL"
                echo "Successful: $SUCCESS ($SUCCESS_RATE%)"
                echo "Failed: $FAILURE"
                echo "✓ Build history fetched"
            else
                echo "⚠️  glab CLI not installed. Install: https://gitlab.com/gitlab-org/cli"
                exit 1
            fi
            ;;

        circleci)
            # Use CircleCI API
            if [ ! -z "$CIRCLE_TOKEN" ]; then
                echo "Fetching last $limit CircleCI builds..."
                PROJECT_SLUG=$(git remote get-url origin | sed 's/.*github.com[:/]\(.*\)\.git/\1/')

                curl -s "https://circleci.com/api/v2/project/github/$PROJECT_SLUG/pipeline?limit=$limit" \
                    -H "Circle-Token: $CIRCLE_TOKEN" > /tmp/ci_builds.json

                echo "✓ Build history fetched"
            else
                echo "⚠️  CIRCLE_TOKEN not set. Set environment variable."
                exit 1
            fi
            ;;

        *)
            echo "⚠️  Build history fetching not implemented for $platform"
            echo "Manual analysis required"
            ;;
    esac

    echo ""
}

fetch_build_history "$CI_PLATFORM" 50

Phase 3: Success Rate Analysis

I'll analyze build success trends over time:

#!/bin/bash
# Analyze build success rates and trends

analyze_success_rates() {
    echo "=== Success Rate Analysis ==="
    echo ""

    if [ ! -f "/tmp/ci_builds.json" ]; then
        echo "⚠️  No build data available"
        return
    fi

    # Overall statistics
    echo "Overall Statistics:"
    TOTAL=$(jq length /tmp/ci_builds.json)
    SUCCESS=$(jq '[.[] | select(.conclusion=="success" or .status=="success")] | length' /tmp/ci_builds.json)
    FAILURE=$(jq '[.[] | select(.conclusion=="failure" or .status=="failed")] | length' /tmp/ci_builds.json)
    IN_PROGRESS=$(jq '[.[] | select(.conclusion=="in_progress" or .status=="running")] | length' /tmp/ci_builds.json)

    SUCCESS_RATE=$(echo "scale=2; $SUCCESS * 100 / $TOTAL" | bc)

    echo "  Total builds: $TOTAL"
    echo "  Successful: $SUCCESS ($SUCCESS_RATE%)"
    echo "  Failed: $FAILURE"
    echo "  In progress: $IN_PROGRESS"
    echo ""

    # Trend analysis (last 10 vs previous 40)
    echo "Trend Analysis:"
    RECENT_SUCCESS=$(jq '[.[:10] | .[] | select(.conclusion=="success" or .status=="success")] | length' /tmp/ci_builds.json)
    RECENT_RATE=$(echo "scale=2; $RECENT_SUCCESS * 100 / 10" | bc)

    PREVIOUS_SUCCESS=$(jq '[.[10:] | .[] | select(.conclusion=="success" or .status=="success")] | length' /tmp/ci_builds.json)
    PREVIOUS_TOTAL=$((TOTAL - 10))
    PREVIOUS_RATE=$(echo "scale=2; $PREVIOUS_SUCCESS * 100 / $PREVIOUS_TOTAL" | bc)

    echo "  Last 10 builds: $RECENT_RATE%"
    echo "  Previous builds: $PREVIOUS_RATE%"

    TREND_DIFF=$(echo "scale=2; $RECENT_RATE - $PREVIOUS_RATE" | bc)
    if (( $(echo "$TREND_DIFF > 0" | bc -l) )); then
        echo "  Trend: ✓ Improving (+$TREND_DIFF%)"
    elif (( $(echo "$TREND_DIFF < 0" | bc -l) )); then
        echo "  Trend: ⚠️  Declining ($TREND_DIFF%)"
    else
        echo "  Trend: Stable"
    fi
    echo ""

    # Success rate by workflow/job
    echo "Success Rate by Workflow:"
    jq -r '.[] | "\(.name // "unknown")|\(.conclusion // .status)"' /tmp/ci_builds.json | \
        awk -F'|' '{workflows[$1]++; if($2=="success") success[$1]++}
                    END {for(w in workflows) printf "  %s: %.0f%% (%d/%d)\n", w, (success[w]/workflows[w])*100, success[w], workflows[w]}' | \
        sort -t: -k2 -n
    echo ""
}

analyze_success_rates

Phase 4: Flaky Test Detection

I'll identify tests that fail inconsistently:

#!/bin/bash
# Detect flaky tests from CI logs

detect_flaky_tests() {
    echo "=== Flaky Test Detection ==="
    echo ""

    # Download recent failed build logs
    echo "Analyzing failed builds for test failures..."

    # Extract test failures from logs (GitHub Actions)
    if [ "$CI_PLATFORM" = "github-actions" ]; then
        # Get failed run IDs
        FAILED_RUNS=$(jq -r '.[] | select(.conclusion=="failure") | .databaseId' /tmp/ci_builds.json | head -20)

        # Track test failures across runs
        > /tmp/test_failures.txt

        for run_id in $FAILED_RUNS; do
            echo "Checking run $run_id..."

            # Download logs and extract test failures
            gh run view "$run_id" --log 2>/dev/null | \
                grep -E "(FAIL|FAILED|Error:|AssertionError|Test failed)" | \
                grep -oP '(test_\w+|it\(["\x27][^\)]+|describe\(["\x27][^\)]+)' >> /tmp/test_failures.txt || true
        done

        if [ -s /tmp/test_failures.txt ]; then
            echo ""
            echo "Test Failure Frequency:"

            # Count occurrences and identify flaky tests
            sort /tmp/test_failures.txt | uniq -c | sort -rn | head -20 | while read count test; do
                # Tests that fail sometimes (not always) are flaky
                TOTAL_RUNS=$(echo "$FAILED_RUNS" | wc -l)
                FAILURE_RATE=$(echo "scale=2; $count * 100 / $TOTAL_RUNS" | bc)

                if (( $(echo "$count > 1 && $FAILURE_RATE < 80" | bc -l) )); then
                    echo "  ⚠️  FLAKY: $test (failed $count/$TOTAL_RUNS times, $FAILURE_RATE%)"
                else
                    echo "  $test (failed $count times)"
                fi
            done

            echo ""
            echo "Recommendations:"
            echo "- Investigate flaky tests marked with ⚠️"
            echo "- Check for race conditions, timing dependencies"
            echo "- Review test isolation and cleanup"
            echo "- Consider retry logic or increased timeouts"
        else
            echo "✓ No test failures detected in recent runs"
        fi
    else
        echo "⚠️  Flaky test detection not yet implemented for $CI_PLATFORM"
        echo "Manual log analysis required"
    fi

    echo ""
}

detect_flaky_tests

Phase 5: Performance Trend Analysis

I'll track build duration and performance:

#!/bin/bash
# Analyze build performance and duration trends

analyze_build_performance() {
    echo "=== Build Performance Analysis ==="
    echo ""

    if [ ! -f "/tmp/ci_builds.json" ]; then
        echo "⚠️  No build data available"
        return
    fi

    # Calculate build durations
    echo "Build Duration Statistics:"

    # Extract durations (GitHub Actions)
    if [ "$CI_PLATFORM" = "github-actions" ]; then
        jq -r '.[] | "\(.createdAt)|\(.updatedAt)"' /tmp/ci_builds.json | while IFS='|' read created updated; do
            if [ ! -z "$created" ] && [ ! -z "$updated" ]; then
                START=$(date -d "$created" +%s 2>/dev/null || date -j -f "%Y-%m-%dT%H:%M:%S" "$created" +%s 2>/dev/null || echo 0)
                END=$(date -d "$updated" +%s 2>/dev/null || date -j -f "%Y-%m-%dT%H:%M:%S" "$updated" +%s 2>/dev/null || echo 0)
                DURATION=$((END - START))
                echo $DURATION
            fi
        done > /tmp/build_durations.txt

        if [ -s /tmp/build_durations.txt ]; then
            # Calculate statistics
            AVG_DURATION=$(awk '{sum+=$1; count++} END {printf "%.0f", sum/count}' /tmp/build_durations.txt)
            MIN_DURATION=$(sort -n /tmp/build_durations.txt | head -1)
            MAX_DURATION=$(sort -n /tmp/build_durations.txt | tail -1)

            # Convert to human readable
            echo "  Average: $(($AVG_DURATION / 60))m $(($AVG_DURATION % 60))s"
            echo "  Fastest: $(($MIN_DURATION / 60))m $(($MIN_DURATION % 60))s"
            echo "  Slowest: $(($MAX_DURATION / 60))m $(($MAX_DURATION % 60))s"

            # Trend analysis
            RECENT_AVG=$(head -10 /tmp/build_durations.txt | awk '{sum+=$1; count++} END {printf "%.0f", sum/count}')
            PREVIOUS_AVG=$(tail -n +11 /tmp/build_durations.txt | awk '{sum+=$1; count++} END {printf "%.0f", sum/count}')

            echo ""
            echo "Performance Trend:"
            echo "  Last 10 builds: $(($RECENT_AVG / 60))m $(($RECENT_AVG % 60))s"
            echo "  Previous builds: $(($PREVIOUS_AVG / 60))m $(($PREVIOUS_AVG % 60))s"

            DIFF=$((RECENT_AVG - PREVIOUS_AVG))
            if [ $DIFF -gt 30 ]; then
                echo "  Trend: ⚠️  Slowing down (+$(($DIFF / 60))m $(($DIFF % 60))s)"
                echo ""
                echo "  Recommendations:"
                echo "  - Review recently added dependencies"
                echo "  - Check for increased test count"
                echo "  - Consider caching strategies"
                echo "  - Profile slow steps"
            elif [ $DIFF -lt -30 ]; then
                echo "  Trend: ✓ Improving (-$((-$DIFF / 60))m $((-$DIFF % 60))s)"
            else
                echo "  Trend: Stable"
            fi
        fi
    fi

    echo ""
}

analyze_build_performance

Phase 6: Failure Pattern Analysis

I'll identify common failure patterns:

#!/bin/bash
# Analyze failure patterns and root causes

analyze_failure_patterns() {
    echo "=== Failure Pattern Analysis ==="
    echo ""

    if [ ! -f "/tmp/ci_builds.json" ]; then
        echo "⚠️  No build data available"
        return
    fi

    FAILED_COUNT=$(jq '[.[] | select(.conclusion=="failure" or .status=="failed")] | length' /tmp/ci_builds.json)

    if [ $FAILED_COUNT -eq 0 ]; then
        echo "✓ No recent failures detected"
        echo ""
        return
    fi

    echo "Analyzing $FAILED_COUNT failed builds..."

    # Common failure categories
    > /tmp/failure_categories.txt

    # Get failed run IDs and analyze logs
    FAILED_RUNS=$(jq -r '.[] | select(.conclusion=="failure" or .status=="failed") | .databaseId // .id' /tmp/ci_builds.json | head -10)

    for run_id in $FAILED_RUNS; do
        if [ "$CI_PLATFORM" = "github-actions" ]; then
            gh run view "$run_id" --log 2>/dev/null | while read line; do
                # Categorize failures
                if echo "$line" | grep -qi "timeout\|timed out"; then
                    echo "timeout" >> /tmp/failure_categories.txt
                elif echo "$line" | grep -qi "out of memory\|oom"; then
                    echo "memory" >> /tmp/failure_categories.txt
                elif echo "$line" | grep -qi "network error\|connection refused\|ECONNREFUSED"; then
                    echo "network" >> /tmp/failure_categories.txt
                elif echo "$line" | grep -qi "npm ERR\|pip error\|cargo error"; then
                    echo "dependency" >> /tmp/failure_categories.txt
                elif echo "$line" | grep -qi "test.*fail\|assertion"; then
                    echo "test" >> /tmp/failure_categories.txt
                elif echo "$line" | grep -qi "lint\|format"; then
                    echo "lint" >> /tmp/failure_categories.txt
                fi
            done
        fi
    done

    if [ -s /tmp/failure_categories.txt ]; then
        echo ""
        echo "Failure Categories:"
        sort /tmp/failure_categories.txt | uniq -c | sort -rn | while read count category; do
            case "$category" in
                timeout)
                    echo "  ⏱️  Timeout issues: $count occurrences"
                    echo "     → Consider increasing timeout limits"
                    ;;
                memory)
                    echo "  💾 Memory issues: $count occurrences"
                    echo "     → Review memory-intensive operations"
                    ;;
                network)
                    echo "  🌐 Network issues: $count occurrences"
                    echo "     → Add retry logic for network calls"
                    ;;
                dependency)
                    echo "  📦 Dependency issues: $count occurrences"
                    echo "     → Lock dependency versions"
                    ;;
                test)
                    echo "  🧪 Test failures: $count occurrences"
                    echo "     → Review test stability"
                    ;;
                lint)
                    echo "  ✨ Lint/format issues: $count occurrences"
                    echo "     → Run linter before commit"
                    ;;
            esac
        done
    fi

    echo ""
}

analyze_failure_patterns

Phase 7: Comprehensive Report

I'll generate a comprehensive monitoring report:

#!/bin/bash
# Generate comprehensive pipeline monitoring report

generate_monitoring_report() {
    echo "========================================"
    echo "CI/CD PIPELINE MONITORING REPORT"
    echo "========================================"
    echo ""
    echo "Generated: $(date)"
    echo "Platform: $CI_PLATFORM"
    echo "Analysis Period: Last 50 builds"
    echo ""

    # Summary from previous analyses
    if [ -f "/tmp/ci_builds.json" ]; then
        TOTAL=$(jq length /tmp/ci_builds.json)
        SUCCESS=$(jq '[.[] | select(.conclusion=="success" or .status=="success")] | length' /tmp/ci_builds.json)
        FAILURE=$(jq '[.[] | select(.conclusion=="failure" or .status=="failed")] | length' /tmp/ci_builds.json)
        SUCCESS_RATE=$(echo "scale=2; $SUCCESS * 100 / $TOTAL" | bc)

        echo "HEALTH SCORE: $SUCCESS_RATE%"

        if (( $(echo "$SUCCESS_RATE >= 90" | bc -l) )); then
            echo "Status: ✓ HEALTHY"
        elif (( $(echo "$SUCCESS_RATE >= 70" | bc -l) )); then
            echo "Status: ⚠️  NEEDS ATTENTION"
        else
            echo "Status: ❌ CRITICAL"
        fi

        echo ""
        echo "KEY METRICS:"
        echo "  Success rate: $SUCCESS_RATE%"
        echo "  Total builds: $TOTAL"
        echo "  Failed builds: $FAILURE"

        if [ -f "/tmp/build_durations.txt" ]; then
            AVG_DURATION=$(awk '{sum+=$1; count++} END {printf "%.0f", sum/count}' /tmp/build_durations.txt)
            echo "  Avg duration: $(($AVG_DURATION / 60))m $(($AVG_DURATION % 60))s"
        fi

        echo ""
        echo "RECOMMENDATIONS:"

        if [ $FAILURE -gt $((TOTAL / 4)) ]; then
            echo "  ⚠️  High failure rate - investigate failing builds"
        fi

        if [ -f "/tmp/test_failures.txt" ] && [ -s "/tmp/test_failures.txt" ]; then
            FLAKY_COUNT=$(sort /tmp/test_failures.txt | uniq -c | awk '$1 > 1 && $1 < 15 {count++} END {print count+0}')
            if [ $FLAKY_COUNT -gt 0 ]; then
                echo "  ⚠️  $FLAKY_COUNT flaky tests detected - see details above"
            fi
        fi

        echo ""
    fi

    echo "========================================"
}

generate_monitoring_report

Integration with Other Skills

Workflow Integration:

After failed builds → /debug-systematic
Before releases → /release-automation (check build health)
During development → /test (local testing)
For CI setup → /ci-setup

Skill Suggestions:

High failure rate → /test-coverage, /test-antipatterns
Flaky tests found → /test-async
Performance degradation → /bundle-analyze, /lighthouse

Practical Examples

Monitor default platform:

/pipeline-monitor              # Auto-detect and analyze

Specific platform:

/pipeline-monitor github       # GitHub Actions
/pipeline-monitor gitlab       # GitLab CI
/pipeline-monitor circle       # CircleCI

Custom time range:

/pipeline-monitor --last 100   # Last 100 builds
/pipeline-monitor --days 7     # Last 7 days

Focus on specific metrics:

/pipeline-monitor --flaky      # Focus on flaky test detection
/pipeline-monitor --performance # Focus on performance trends

What Gets Analyzed

Metrics Tracked:

Build success/failure rates
Build duration and trends
Flaky test identification
Failure pattern analysis
Performance degradation
Workflow-specific metrics

Platforms Supported:

GitHub Actions (via gh CLI)
GitLab CI (via glab CLI)
CircleCI (via API with token)
Jenkins (manual log analysis)
Travis CI (basic support)
Azure Pipelines (basic support)

Safety Guarantees

What I'll NEVER do:

Modify CI/CD configuration files
Trigger builds or deployments
Cancel running builds
Delete build history
Modify workflow settings

What I WILL do:

Read-only analysis of build data
Statistical trend analysis
Actionable recommendations
Clear reporting with context

Credits

This skill integrates:

GitHub CLI - GitHub Actions integration
GitLab CLI - GitLab CI integration
CircleCI API - CircleCI integration
Statistical Analysis - Trend detection algorithms

Token Budget

Target: 2,000-3,500 tokens per execution

Phase 1-2: ~600 tokens (detection, fetch)
Phase 3-4: ~800 tokens (success rates, flaky tests)
Phase 5-6: ~800 tokens (performance, patterns)
Phase 7: ~500 tokens (reporting)

Optimization Strategy:

Platform detection via bash (no file reading)
API/CLI calls for build data (minimal tokens)
Statistical analysis without full log parsing
Grep for specific error patterns
Summary-based reporting

This ensures comprehensive pipeline monitoring while maintaining efficiency and token budget compliance.