devops-tools

Monitor background processes from Claude Code using sentinel files, heartbeat liveness, and subagent polling. Best practices and anti-patterns for autonomous loops that need to kick off work, detect completion/failure/hang/timeout, and resume the main context without wasting tokens. TRIGGERS - monitor background process, sentinel file, heartbeat monitoring, process supervision, agentic loop monitor, background task health, detect hung process, poll for completion, watchdog pattern, process liveness, monitor long-running task, agent poll loop, circuit breaker pattern.

Claude Code OAuth proxy patterns and anti-patterns for multi-provider model routing. TRIGGERS - proxy Claude Code, OAuth token Keychain, route Haiku to MiniMax, ANTHROPIC_BASE_URL, model routing proxy, claude-code-proxy, proxy-toggle, multi-provider setup, anthropic-beta oauth, proxy auth failure, go proxy, failover proxy, launchd proxy, proxy failover

ClickHouse Cloud user and permission management. TRIGGERS - create ClickHouse user, ClickHouse permissions, ClickHouse Cloud credentials.

Generate DBeaver config from Pydantic ClickHouse models. TRIGGERS - DBeaver config, ClickHouse connection, database client config.

Deploy static HTML files to Cloudflare Workers with 1Password credential management. TRIGGERS - Cloudflare Workers deploy, publish static site, wrangler deploy, static hosting, cloudflare publish, CF Workers, HTML hosting, workers.dev, static assets deploy

macOS disk cleanup, cache pruning, stale file detection, and Downloads triage. TRIGGERS - disk space, cleanup, disk usage, stale files, cache clean, brew cleanup, forgotten files, Downloads cleanup, free space, storage, dust, dua, gdu, ncdu.

Concurrency safety patterns for distributed pueue + mise + systemd-run job pipelines. TRIGGERS - queue pueue jobs, deploy to remote host, concurrent job collisions, checkpoint races, resource guards, cgroup memory limits, systemd-run, autoscale, batch processing safety, job parameter isolation.

Validate and test Doppler secrets. TRIGGERS - add to Doppler, store secret, validate token, test credentials.

Manage credentials and secrets through Doppler for publishing and deployment workflows. Use whenever the user needs to publish Python packages to PyPI, rotate AWS credentials, manage Doppler secrets, or configure credential pipelines for CI/CD. Do NOT use for 1Password vault operations or for secrets that are not managed through Doppler.

Dual-channel notifications on watchexec events. TRIGGERS - watchexec alerts, Telegram+Pushover, file change notifications.

Programmatic Firecrawl usage, self-hosted operations, academic paper routing, recursive deep research, and raw corpus persistence. TRIGGERS - firecrawl search, firecrawl scrape, academic paper, arxiv, deep research, recursive search, research pattern, corpus persistence, firecrawl, self-hosted scraping, web scrape, scraper wrapper, littleblack, ZeroTier scraping.

Configure a MacBook as an always-on-AC desktop workstation with USB device resilience, battery longevity, and self-healing audio device monitoring. Covers charge limits, sleep optimization, powered USB hub with uhubctl, and the AudioDeviceMonitor Swift guardian (state machine + wake detection + heartbeat + recovery cascade + Telegram notification). Use whenever the user mentions desktop mode, always on AC, charge limit, battery longevity, USB device disappearing, sleep settings, powered hub, thermal management, or USB microphone dead after sleep. Do NOT use for general macOS troubleshooting unrelated to the desktop workstation pattern.

Patterns for efficient ML data pipelines using Polars, Arrow, and ClickHouse. TRIGGERS - data pipeline, polars vs pandas, arrow format, clickhouse ml, efficient loading, zero-copy, memory optimization.

POC validation patterns to catch issues before committing to long-running ML experiments. TRIGGERS - fail-fast, POC validation, preflight check, experiment validation, schema validation, gradient check, sanity check, smoke test.

MLflow experiment tracking via Python API. TRIGGERS - MLflow metrics, log backtest, experiment tracking, search runs.

Migrate Claude Code project sessions when renaming directories. TRIGGERS - directory rename, move project, migrate sessions, project path change, workspace reorganization, rename folder.

Manage long-running jobs and batch processing with pueue queue orchestration, CLI telemetry, and companion monitoring tools (noti, ntfy, mprocs, task-spooler). Use whenever the user wants to queue a job, run tasks on bigblack or littleblack, manage long-running processes, do batch processing, set up pueue callbacks or priorities, get notified when jobs finish (noti/ntfy), watch multiple processes (mprocs), or needs GPU workstation job management. Also use for cache population tasks that run in the background. Do NOT use for simple foreground shell commands that complete quickly or for job scheduling via cron/launchd.

Python logging with loguru and platformdirs. TRIGGERS - loguru, structured logging, JSONL logs, log rotation, XDG directories.

Memory-safe Python script patterns for long-running processes under systemd MemoryMax constraints. Covers allocator purge (mimalloc/glibc malloc_trim), HTTP response lifecycle, DataFrame cleanup, thread-local connection reuse, and periodic GC cadence. Battle-tested through 5 OOM optimization cycles on production GPU workstations. Use this skill proactively whenever writing or reviewing Python scripts that: run under systemd with MemoryMax, process data in loops (downloads, ETL, backfill), use ThreadPoolExecutor, or make repeated HTTP requests. Also use when diagnosing OOM kills, RSS creep, or fd exhaustion in Python services. TRIGGERS - memory optimization, OOM prevention, RSS reduction, malloc_trim, systemd MemoryMax, memory leak, allocator purge, memory-safe script, RSS creep, fd exhaustion, SIGKILL status 9, MemoryHigh, glibc arena, mimalloc purge, requests memory leak, ThreadPoolExecutor cleanup.

Session log provenance tracking. TRIGGERS - who created, trace origin, session archaeology, ADR reference.

Analyze Claude Code sessions in three expert modes — Handoff Document (exhaustive context extraction for the next developer or session), Error Forensics (complete inventory of warnings and errors Claude ignored or deferred), and Chronological Summary (dense technical timeline with key outcomes). Use when the user wants a session handoff, asks what happened, wants to know what errors were missed or ignored, requests a retrospective or debrief, mentions session history or session analysis, or wants to understand past work in the current repo. Do not use for live debugging or code review.

Diagnose and recover Claude Code session issues including missing conversations, corrupted sessions, and session file problems. Use whenever the user sees 'No conversations found to resume', sessions appear in wrong locations, or session files are missing or corrupted. Do NOT use for general Claude Code configuration or for issues unrelated to session persistence and recovery.

Create alpha-forge git worktrees with auto branch naming. TRIGGERS - create worktree, new worktree, alpha-forge worktree.

1 hook across 1 event