Skip to content

Findings · Updated 2026-10-29

Repositories of practical value

Selected by task, documentation, tests and maintenance — not by stars.

Selection · 2026-10-29

20 repositories → 18 practices across four tracks: DevOps, Platform Engineering / DevSecOps, MLOps / LLMOps and AI tools for engineers. Ansible-lint with Molecule and promptfoo with garak are merged.

P1 — immediately useful for the VPS and GitLab CE; P2 — when the matching CI, LLM workload or security tasks exist; P3 — the next step on the career track.

Priority Track Repository What to put to work Status
P1 DevOps renovatebot/renovate Automatic dependency and image updates in GitLab Not studied
P1 DevOps firecow/gitlab-ci-local Run .gitlab-ci.yml locally before push Not studied
P1 Platform/DevSecOps ossf/scorecard Assess repository security and maintenance Not studied
P1 AI tools snyk/agent-scan Check MCP servers and skills before installing Not studied
P2 DevOps ansible/ansible-lint Static analysis of Ansible roles Not studied
P2 DevOps ansible/molecule Testing Ansible roles: converge, idempotence, verify Not studied
P2 DevOps crowdsecurity/crowdsec Protect a public VPS using Nginx and SSH logs Not studied
P2 Platform/DevSecOps zizmorcore/zizmor Security audit of GitHub Actions Not studied
P2 Platform/DevSecOps anchore/grype Find CVEs in a Syft SBOM Not studied
P2 MLOps/LLMOps BerriAI/litellm A single LLM gateway: keys, budgets, limits, fallback Not studied
P2 MLOps/LLMOps langfuse/langfuse LLM observability: traces, tokens, cost, evals Not studied
P2 MLOps/LLMOps promptfoo/promptfoo Regression tests and red teaming of LLM apps in CI Not studied
P2 MLOps/LLMOps NVIDIA/garak Model vulnerability scanner: injection, leaks, jailbreak Not studied
P2 MLOps/LLMOps DataTalksClub/mlops-zoomcamp A structured MLOps course for moving over from DevOps Not studied
P2 AI tools anthropics/skills Agent Skills templates for your own procedures Not studied
P3 Platform/DevSecOps backstage/backstage Internal developer portal (IDP) Not studied
P3 MLOps/LLMOps vllm-project/vllm Serving LLMs on your own hardware Not studied
P3 AI tools containers/kubernetes-mcp-server Restricted AI-agent access to Kubernetes Not studied
P3 AI tools robusta-dev/holmesgpt AI analysis of Prometheus/Alertmanager alerts Not studied
P3 AI tools msitarzewski/agency-agents Role scaffolds for subagents Not studied

Trivy: pin the image by digest

In March 2026 Trivy was compromised: the malicious v0.69.4 release, images v0.69.5–v0.69.6 and rewritten trivy-action and setup-trivy tags. Through an unpinned Trivy in the LiteLLM project's CI, a publishing token was stolen, and malicious litellm 1.82.7 and 1.82.8 were published to PyPI. The Trivy in CI practice: image by digest only, no latest. Sources: CSA, SafeDep, Snyk.

What to work through next

  • Renovate: a scheduled pipeline in GitLab CE, images by digest, one update group per week.
  • gitlab-ci-local: run the pipeline locally before push.
  • Scorecard: assess the security and maintenance of the repositories in the selections.
  • agent-scan: check every skill and MCP server before installing.

AI tool catalogs

  • agency-agents — role scaffolds for subagents, not proven practices: CI checks the form, not the behavior.
  • awesome-ai-plugins — only for finding candidates: trust scores are assigned by the catalog owner's own scanner, and some entries are its own products.

Selection · 2026-10-11

12 repositories → 11 practices: Actionlint and ShellCheck are merged into one. All practices in this selection are not studied yet.

P1 — useful for the current VPS; P2 — when the matching CI, Kubernetes or security tasks exist; P3 — the next review once requirements are defined.

Priority Track Repository What to put to work Status
P1 DevOps restic/restic Restoring VPS data Not studied
P1 DevOps healthchecks/healthchecks Detecting a missed cron job Not studied
P1 DevOps prometheus/blackbox_exporter Checking the site and TLS Not studied
P1 DevOps goss-org/goss VPS acceptance after changes Not studied
P2 DevOps rhysd/actionlint Checking GitHub Actions Not studied
P2 DevOps koalaman/shellcheck Checking shell scripts Not studied
P2 DevOps yannh/kubeconform Checking Kubernetes YAML Not studied
P2 Platform/DevSecOps anchore/syft Image component inventory Not studied
P2 Platform/DevSecOps getsops/sops Encrypted configs in Git Not studied
P3 Platform/DevSecOps open-policy-agent/conftest Configuration policies Not studied
P3 DevOps slok/sloth SLOs and error budget burn Not studied
P3 Platform/DevSecOps sigstore/cosign Verifying image provenance Not studied

What to work through next

  • Conftest: one configuration policy with positive/negative tests.
  • Sloth: one SLO, verified SLIs and error budget burn rules.
  • Cosign: verify the signature against a trusted identity/issuer or key and the image digest.

Archive · 2026-04-15

Priority Track Repository What to put to work Status
— DevOps kamranahmedse/developer-roadmap A map of DevOps areas and gaps Studied
— DevOps trimstray/the-book-of-secret-knowledge Linux diagnostics, networking and hardening Studied
— DevOps bregman-arie/devops-exercises Hands-on exercises and worked problems Studied
— DevOps binhnguyennus/awesome-scalability Reliability and scaling Studied
— DevOps kelseyhightower/kubernetes-the-hard-way How Kubernetes works Studied
— DevOps veggiemonk/awesome-docker Docker tools and practices Studied
— DevOps MichaelCade/90DaysOfDevOps A step-by-step learning program Studied
— DevOps ramitsurana/awesome-kubernetes Navigating the Kubernetes ecosystem Studied
— DevOps dastergon/awesome-sre SRE, SLOs and incident handling Studied
— DevOps bregman-arie/devops-resources Additional DevOps materials Studied