Personal collection of agent skills for use with Claude Code and other LLM agents. Each skill is a reusable instruction set that primes the agent with domain knowledge, workflow patterns, and failure playbooks — so you don't have to re-explain the same context every session.
Catalog: https://skills.sh/andreab67/agent-skills
| Skill | Docs | Description |
|---|---|---|
postgres-ops |
docs | Production PostgreSQL operations: incident diagnosis, query plan analysis, HA/DR design, migrations, connection pooling, observability, and federal/DoD hardening (STIG, pgaudit, RLS). |
nextjs-monorepo-ci |
docs | GitLab CI/CD for Next.js monorepos: six-stage pipeline (validate → security → build → obfuscate → package → notify), Kaniko unprivileged builds, Semgrep, gitleaks, javascript-obfuscator, Harbor registry, IndexNow post-deploy ping. |
k8s-nextjs-deploy |
docs | Kubernetes deployment patterns for Next.js apps: Deployment/Service/Ingress manifests, Harbor pull secret rotation, Traefik + cert-manager TLS, Linkerd service mesh, multi-app namespaces, failure diagnosis (ImagePullBackOff, CreateContainerConfigError). |
confluence-to-nextjs |
docs | Migrate Atlassian Confluence pages to a self-hosted Next.js knowledge base: fetch via REST API v2, convert storage-format HTML to semantic JSX, anchor navigation, sticky table of contents, and Vitest test patterns. |
ubuntu24-stig |
docs | DISA STIG V1R1 hardening for Ubuntu 24.04 LTS on AWS EC2: OpenSCAP scanning with correct benchmark IDs, idempotent remediation shell script for MAC-2_Sensitive, auditd rules, SSH FIPS 140-3 ciphers, PAM lockout/pwquality, AIDE integrity monitoring, and AWS skip list. |
login-gov |
docs | Integrate applications with login.gov federal identity provider: OIDC authorization code flow, private_key_jwt client authentication, ACR value selection (auth-only, IAL2, facial-match, PIV/CAC), user attribute scopes, sandbox setup, token validation, and Partner Portal configuration. |
arcgis-enterprise-k8s |
docs | Deploy and operate ArcGIS Enterprise on Kubernetes 12.0: deployment profiles (Dev/Standard/Enhanced HA), EKS/AKS/GKE/OpenShift sizing, persistent volume planning, L4/L7 load balancer configuration, TLS requirements, upgrade procedures, and pod failure diagnosis. |
loki-best-practices |
docs | Production Grafana Loki operations: reason-keyed ingest-discard triage (rate_limited, stream_limit, out_of_order, invalid_labels), WAL replay and chunk-flush recovery, Helm deployment-mode and schema-period pitfalls, OTel/Promtail/Alloy push debugging, query-performance tuning (max_outstanding_requests, cardinality, regex rewrites), S3/IRSA and compactor retention, ring health and HA rolling upgrades, and the seven alerts every Loki cluster should have. |
session-handoff |
docs | Rescue critical session context to persistent auto-memory before /clear, /compact, or any conversation wipe. One-sentence rescue test (would re-reading the repo recover this?), six categories mapped to user/feedback/project/reference, hard exclusions for secrets and code-derivable facts, consistent slug-prefix tagging for retrieval, review-before-write via AskUserQuestion, conflict resolution against stale entries, anti-patterns gallery, and a worked end-to-end example. Optional UserPromptSubmit hook intercepts /clear for true auto-activation. |
anthropic-sdk |
docs | Anthropic Python SDK expert: Messages API, streaming, multi-turn tool use loop, count_tokens() pre-flight, prompt caching with cache_control, vision input, model selection (Opus 4.7 / Sonnet 4.6 / Haiku 4.5) with pricing table and cost estimation helper. |
openai-sdk |
docs | OpenAI Python SDK expert: chat completions, streaming, function calling with tool_call loop, JSON structured output, embeddings, tiktoken token counting with count_message_tokens() helper, GPT-4o/mini/o1/o3 pricing table, and cost estimation. |
kilo-gateway |
docs | Kilo.ai OpenAI-compatible gateway over 500+ models: base_url="https://api.kilo.ai/api/gateway", BYOK setup (zero Kilo markup), provider/model-name routing format, Python/TypeScript/cURL/LangChain patterns, prompt-cache header, and free-model :free suffix. |
openrouter |
docs | OpenRouter unified API over 300+ models: OpenAI-compatible chat/streaming/JSON-mode, model fallbacks and openrouter/auto routing, sampling-parameter guide, plus account/analytics endpoints for dashboards — /credits, /keys, /activity (management-key-only, last-30-completed-UTC-days lag), beta /analytics/query, and the /models pricing catalog (full field reference). |
magnific |
docs | Magnific generative-media API (api.magnific.com, part of Freepik): the submit→task_id→poll-or-webhook async task model, x-magnific-api-key auth, and Svix-style HMAC webhook verification — across ~60 endpoints. Image generation (Mystic, Flux ×7, Seedream, Z-Image, Runway), the flagship Creative/Precision upscalers plus relight/style-transfer/remove-bg/expand, video (Kling, MiniMax, WAN, Runway, LTX, Seedance, PixVerse, OmniHuman, VFX), audio (music/SFX/isolation), and team analytics + stock (icons/videos/templates). Full per-endpoint catalog in references/, rate limits, credit pricing, and the MCP server. |
code-review |
docs | Adversarially-verified multi-angle code review: ten angles (five bug-hunting, five quality) fanned to cheap subagents, every candidate finding refuted against the actual code, a coverage-gap sweep, and a severity-ranked report of confirmed findings only — each with file:line and a concrete failure scenario. "No confirmed findings" is a valid result. Portable across Claude Code, hermes-agent, Kilo Code, and Codex; no-subagent hosts (Cursor/AnythingLLM) run the angles as sequential focused passes. Host mechanics in references/. |
Install a single skill globally:
npx skills add andreab67/agent-skills@postgres-ops -g -y
npx skills add andreab67/agent-skills@nextjs-monorepo-ci -g -y
npx skills add andreab67/agent-skills@k8s-nextjs-deploy -g -y
npx skills add andreab67/agent-skills@confluence-to-nextjs -g -y
npx skills add andreab67/agent-skills@ubuntu24-stig -g -y
npx skills add andreab67/agent-skills@login-gov -g -y
npx skills add andreab67/agent-skills@arcgis-enterprise-k8s -g -y
npx skills add andreab67/agent-skills@loki-best-practices -g -y
npx skills add andreab67/agent-skills@session-handoff -g -y
npx skills add andreab67/agent-skills@anthropic-sdk -g -y
npx skills add andreab67/agent-skills@openai-sdk -g -y
npx skills add andreab67/agent-skills@kilo-gateway -g -y
npx skills add andreab67/agent-skills@openrouter -g -y
npx skills add andreab67/agent-skills@magnific -g -y
npx skills add andreab67/agent-skills@code-review -g -yOr install all skills at once:
for skill in postgres-ops nextjs-monorepo-ci k8s-nextjs-deploy confluence-to-nextjs ubuntu24-stig login-gov arcgis-enterprise-k8s loki-best-practices session-handoff anthropic-sdk openai-sdk kilo-gateway openrouter magnific code-review; do
npx skills add "andreab67/agent-skills@$skill" -g -y
doneEach skill lives in its own directory with a SKILL.md file. The frontmatter declares:
name— used by the skills CLI for install/referencedescription— the trigger condition that tells the agent when to activate this skill
When a skill is active, the agent reads the full SKILL.md and follows its workflow, terminology, and failure patterns rather than relying on generic training data.
Bug reports, fixes, and new skills welcome — see CONTRIBUTING.md for scope, layout, and PR conventions.
BSD 3-Clause — see LICENSE