Refresh "Claude Skills I Actually Use for DevOps" with an FAQ

[alexleventer]

Freshens engin-diri's post The Claude Skills I Actually Use for DevOps. It's a fast-moving topic, and the changes make the post easier to get answers from.

What changed

• New "Frequently asked questions" section (6 Q&As). Covers the questions readers most often bring to a post like this: the DevOps skills stack worth starting with, skills vs. MCP servers, whether skills are safe to install, which skills help with Kubernetes, Ansible/Azure coverage, and whether skills work outside Claude Code. Each answer is self-contained. Opts into FAQPage schema via faq_schema: true (same mechanism ai-infrastructure-tools uses).
• Clearer meta description. The old one didn't say what the post was about; the new one names it — Claude Code skills for DevOps (Pulumi, Kubernetes, monitoring, security, incident response) and vetting them safely.
• updated: 2026-07-22 so the post shows an "Updated" date.

Integrity notes

• Every FAQ answer is grounded in the post's existing content — no new skills or repos were invented, and the Ansible/Azure answer honestly points to the general devops-engineer skill rather than a fabricated one.
• Engin's voice and authorship are unchanged.
• All external repo/doc links in the post were re-verified live (HTTP 200).

Verification

• lint-markdown passes.
• Local build renders the FAQPage JSON-LD with all 6 questions, the new meta description, and the "Updated Jul 22, 2026" line.

Review

Left as draft — this is Engin's post, so it needs his sign-off before going ready. Happy to run it through docs-review first, or adjust any of the FAQ answers to match how he'd phrase them.

🤖 Generated with Claude Code

[github-actions]

Pre-merge Review — Last updated 2026-07-22T21:59:08Z

Tip

Summary: This PR refreshes an existing blog post ("The Claude Skills I Actually Use for DevOps") by appending a six-entry "Frequently asked questions" section, enabling faq_schema: true for FAQ rich results, rewriting the meta description, and stamping updated: 2026-07-22. The kind of wrongness that would hurt a reader here is an incorrect safety command: the FAQ tells readers to vet a skill by running uvx mcp-scan@latest --skills, and the verification step found the tool appears to have been renamed (to snyk-agent-scan) with the skill-scanning flag semantics changed — a reader who copies it may hit a command that no longer resolves. Two cross-agent capability claims (skills also running in Cursor / GitHub Copilot) couldn't be confirmed because the cited source, agentskills.io, is a JavaScript app whose content didn't load for the verifier. Passes run: FAQ claim verification, editorial-balance, frontmatter/URL sweep, and prose linting. The FAQ's overlap with earlier sections is intentional (standalone, search-landable answers) and not treated as a defect.

Review confidence:

Dimension	Level	Notes
mechanics	HIGH
facts	MEDIUM	One command flagged ❌ contradicted (the mcp-scan tool appears renamed); two cross-agent claims unverifiable because the cited site didn't render for the verifier.
coherence	HIGH
editorial balance	HIGH

Investigation log

• Cross-sibling reads: not run (not in a templated section)
• External claim verification: 4 of 12 claims verified (2 unverifiable, 1 contradicted) · 4 specialists (numerical, cross-reference, capability, framing); 0 cross-specialist corroborations · routed: 0 inline, 7 Pass 1, 2 Pass 2 (verified 0, contradicted 0, unverifiable 2), 3 Pass 3 (verified 1, contradicted 0, unverifiable 2).
• Cited-claim spot-checks: 2 of 2 cited claims fetched and compared
• Frontmatter sweep: ran on body + meta_desc
• Temporal-trigger sweep: ran (recency words present in diff; spot-check in-review)
• Code execution: not run (no static/programs/ change)
• Code-examples checks: not run (no fenced code blocks in content files)
• Editorial-balance pass: ran (15 H2 sections, 1 flags fired)

🚨 Outstanding	⚠️ Low-confidence	💡 Pre-existing	✅ Resolved
1	6	0	0

🔍 Verification trail

15 claims extracted · 4 verified · 2 unverifiable · 1 contradicted

• L4 in content/blog/top-8-claude-skills-devops-2026/index.md "updated: 2026-07-22" → ➖ not-a-claim (evidence: This is a front-matter 'updated' date field, which is metadata reflecting when the author last edited the post, not a falsifiable factual claim to verify against an external source.; source: content/blog/top-8-claude-skills-devops-2026/index.md front matter)
• L4 in content/blog/top-8-claude-skills-devops-2026/index.md "This blog post was updated on 2026-07-22." → ➖ not-a-claim (evidence: The file's frontmatter itself sets updated: 2026-07-22, so this line is a direct restatement of the post's own metadata field, not a falsifiable third-party claim.; source: repo:content/blog/top-8-claude-skills-devops-2026/index.md)
• L6 in content/blog/top-8-claude-skills-devops-2026/index.md "The post describes Claude Code skills a DevOps engineer actually uses for Pulumi, Kubernetes, monitoring, security, and incident response, plus how to vet them…" → ➖ not-a-claim (evidence: This is a self-referential summary of the blog post's own content/scope written by its author, not a falsifiable third-party assertion requiring external verification.; source: content/blog/top-8-claude-skills-devops-2026/index.md)
• L267 in content/blog/top-8-claude-skills-devops-2026/index.md "The recommended DevOps skill stack consists of pulumi-typescript, pulumi-esc, and pulumi-best-practices for infrastructure code, monitoring-expert and kubernet…" → ➖ not-a-claim (evidence: Line 267 is the author's own FAQ-answer summarizing their personal recommended skill stack, and every skill named (pulumi-typescript, pulumi-esc, pulumi-best-practices, monitoring-expert, kubernetes-specialist, k8s-security-policies, secur…; source: repo:content/blog/top-8-claude-skills-devops-2026/index.md)
• L269-271 in content/blog/top-8-claude-skills-devops-2026/index.md "MCP servers give Claude tools (access to external systems like a registry, cloud APIs, or a database), while Claude skills give Claude process (workflows and j…" → ➖ not-a-claim (evidence: This is the blog author's own explanatory analogy/framing distinguishing MCP servers (tool access) from Claude skills (process/workflow guidance) in an FAQ section, not a falsifiable factual assertion about a specific Pulumi product or a t…; source: repo:content/blog/top-8-claude-skills-devops-2026/index.md L269-271)
• L275 in content/blog/top-8-claude-skills-devops-2026/index.md "Running uvx mcp-scan@latest --skills is a way to check the safety of a Claude skill before installing it." → ❌ contradicted (evidence: The invariantlabs-ai/mcp-scan project is now snyk/agent-scan; its README documents the command as uvx snyk-agent-scan@latest (not mcp-scan), and shows skill scanning as automatic/default behavior with a --no-skills flag to *disab…; source: gh api repos/invariantlabs-ai/mcp-scan/contents/README.md (redirects to snyk/agent-scan); gh api repos/invariantlabs-ai/mcp-scan; intuition: The package has been renamed from mcp-scan to snyk-agent-scan and the CLI flag semantics are inverted (--no-skills to d…)
• L275 in content/blog/top-8-claude-skills-devops-2026/index.md "Snyk's ToxicSkills research found that 13.4% of the 3,984 scanned public Claude skills had critical vulnerabilities." → ✅ verified (evidence: Snyk's official blog states: "The findings are stark: 13.4% of all skills, or 534 in total, all contain at least one critical-level security issue" from a scan of "3,984 skills from ClawHub and skills.sh."; source: https://snyk.io/blog/toxicskills-malicious-ai-agent-skills-clawhub/)
• L279 in content/blog/top-8-claude-skills-devops-2026/index.md "The kubernetes-specialist Claude skill produces production-ready manifests with security contexts, resource limits, probes, and pod disruption budgets." → ✅ verified (evidence: The kubernetes-specialist SKILL.md (jeffallan/claude-skills) explicitly instructs: "Set resource requests and limits on all containers", "Include liveness and readiness probes", and provides a full YAML example with `securityContext: runAs…; source: gh api repos/Jeffallan/claude-skills/contents/skills/kubernetes-specialist/SKILL.md)
• L283 in content/blog/top-8-claude-skills-devops-2026/index.md "The general-purpose devops-engineer skill covers CI/CD pipelines and infrastructure as code across AWS, GCP, and Azure." → ✅ verified (evidence: The upstream SKILL.md for jeffallan/claude-skills' devops-engineer skill lists under "When to Use This Skill": "Infrastructure as code (Terraform, Pulumi)" and "Cloud platform configuration (AWS, GCP, Azure)", with description "Creates Doc…; source: gh api repos/Jeffallan/claude-skills/contents/skills/devops-engineer/SKILL.md)
• L283 in content/blog/top-8-claude-skills-devops-2026/index.md "Claude skills follow an open standard, which supports writing custom skills for Ansible or Azure DevOps pipelines." → ✅ verified (evidence: The article itself states at L283: "For Ansible or an Azure DevOps pipeline specifically, the better move is to write your own skill — they follow an open standard — that encodes your team's exact playbook," and elsewhere cites the Agent S…; source: repo:content/blog/top-8-claude-skills-devops-2026/index.md (L283, L287))
• L287 in content/blog/top-8-claude-skills-devops-2026/index.md "A skill written for Claude Code also works in Cursor, GitHub Copilot, and other agents that support the Agent Skills standard." → 🤷 unverifiable (evidence: The pre-fetched body of agentskills.io is truncated navigation/header content and does not contain any statement about cross-agent compatibility (Cursor, GitHub Copilot, etc.) that would confirm or deny the claim.; source: https://agentskills.io)
• L291 in content/blog/top-8-claude-skills-devops-2026/index.md "Skills work in Claude Code, Cursor, GitHub Copilot, and anything else that supports the Agent Skills standard." → 🤷 unverifiable (evidence: The pre-fetched page body is just the site's nav/header shell (documentation index landing page) and doesn't contain the specific list of tools (Claude Code, Cursor, GitHub Copilot) that support the Agent Skills standard, so the claim's sp…; source: https://agentskills.io)
• L0 in content/blog/top-8-claude-skills-devops-2026/index.md "The stack I reach for: pulumi-typescript, pulumi-esc, and pulumi-best-practices for infrastructure code, monitoring-expert and kubernetes-specialist f" → 🚩 flagged (readthrough: self-redundancy)
• L0 in content/blog/top-8-claude-skills-devops-2026/index.md "MCP servers give Claude tools: access to external systems like your registry, cloud APIs, or a database. Skills give Claude process: the workflows and judgment" → 🚩 flagged (readthrough: self-redundancy)
• L0 in content/blog/top-8-claude-skills-devops-2026/index.md "Treat them like any third-party dependency, because they run with your agent's permissions. Snyk's ToxicSkills research scanned 3,984 public skills and found 13" → 🚩 flagged (readthrough: self-redundancy)

📊 Editorial balance

Section depth, mention distribution, recommendation steering

• Section depth: 15 H2 sections (mean 10.4 lines, median 10, std 6.4). Outliers: Putting it together: 31 (3.1× median). The lone outlier is pre-existing body content this PR didn't touch; the new FAQ section (12 lines) sits right at the mean, so no balance flag on the change itself.
• Vendor / entity mentions (new FAQ only): Pulumi 4 (pulumi-typescript, pulumi-esc, pulumi-best-practices, plus "Pulumi") · Kubernetes 3 · Claude Code 3 · Snyk 1 · Cursor 2 · GitHub Copilot 2 · ChatGPT 1 · Ansible 2 · Azure DevOps 2 · AWS/GCP/Azure 1 each · ArgoCD / Flux CD 1 each. Competitor agents are named as first-class, compatible options ("no vendor lock-in"), consistent with the site's AI-positioning rule.
• FAQ steering: 6 FAQ entries; only the "best skills for DevOps" answer leads with Pulumi skills (appropriate for a Pulumi blog); the other five are product-neutral (MCP vs skills, safety, Kubernetes, Ansible/Azure DevOps, cross-agent portability). No over-steering.

🚨 Outstanding in this PR

These must be resolved or refuted before merging.

• [L275] content/blog/top-8-claude-skills-devops-2026/index.md — "Running uvx mcp-scan@latest --skills is a way to check the safety of a Claude skill before installing it." — flagged ❌ contradicted. The verification step found that the mcp-scan project appears to have been renamed: invariantlabs-ai/mcp-scan now redirects to snyk/agent-scan, whose README documents the invocation as uvx snyk-agent-scan@latest — and skill scanning is on by default, disabled with --no-skills rather than enabled with --skills. So the command as written may not resolve, and the --skills flag may not exist. Please confirm against the tool's current README and update the command. Likely rewrite:

  run uvx snyk-agent-scan@latest (skills are scanned by default),

  If the mcp-scan package name and --skills flag are still valid on the date of publish, keep the line as-is and resolve this note — but verify first, since a broken safety command undercuts the section's whole point.

⚠️ Low-confidence

Review each and resolve as appropriate — these don't block the PR.

• [L287] content/blog/top-8-claude-skills-devops-2026/index.md — "A skill written for Claude Code also works in Cursor, GitHub Copilot, and other agents that support the Agent Skills standard." — 🤷 unverifiable. The claim cites agentskills.io, but that site is a JavaScript app and only its nav/header shell loaded for the verifier, so the cross-agent compatibility (Cursor, GitHub Copilot) couldn't be confirmed from the page. Not treated as a blocker — it's a plausible claim about an open standard and it carries a citation. Author check: does agentskills.io (or the standard's own docs) explicitly state Cursor and GitHub Copilot support skills? A deep-link to that statement would make the claim self-verifying for readers.

• [L291] content/blog/top-8-claude-skills-devops-2026/index.md — "Skills work in Claude Code, Cursor, GitHub Copilot, and anything else that supports the Agent Skills standard." — 🤷 unverifiable. Same claim and same cited source as L287 (agentskills.io didn't render its content for the verifier); the two lines restate each other. Not a blocker. Author check: confirm the named agents (Cursor, GitHub Copilot) do support the Agent Skills standard, ideally with a link to where that's documented.

Style findings

Found by pattern-based linting; Findings may be false positives.

• line 265: [style] heading capitalization — Heading 'What are the best Claude skills for DevOps?' should use sentence case (capitalize only the first word and proper nouns).
• line 279: [style] difficulty qualifier — Avoid difficulty qualifier 'just' -- it judges difficulty for the reader (STYLE-GUIDE.md §Inclusive Language).
• line 285: [style] heading capitalization — Heading 'Do Claude skills work outside Claude Code?' should use sentence case (capitalize only the first word and proper nouns).
• line 287: [style] filler — Don't start a sentence with 'There is'.

💡 Pre-existing issues in touched files (optional)

No pre-existing issues in touched files.

✅ Resolved since last review

No items resolved since the last review.

📜 Review history

• 2026-07-22T21:59:08Z — FAQ-refresh PR: flagged one outdated safety command (mcp-scan appears renamed to snyk-agent-scan) as ❌ contradicted; two cross-agent claims unverifiable (cited site didn't render); FAQ/body overlap judged intentional, not flagged. (9520ed0)

---

Need a re-review? Want to dispute a finding? Mention @claude and include #update-review.
(For ad-hoc questions or fixes, just @claude — no hashtag.)

[pulumi-bot]

Your site preview for commit 9520ed0 is ready! 🎉

http://www-testing-pulumi-docs-origin-pr-20459-9520ed0e.s3-website.us-west-2.amazonaws.com

Changed pages:

• /blog/top-8-claude-skills-devops-2026/