hooks — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited hooks (Hook) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
A personal collection of agent skills for software and research engineering, packaged as a single cross-platform plugin. Each skill is a self-contained, model-agnostic instruction set that runs on Claude (Claude Code), Codex, and Antigravity (Gemini) from the same source.
Every skill encodes a disciplined workflow rather than a prompt: classify the task, gather evidence, apply domain references, and verify before handoff. The goal is to keep agent output grounded in explicit forces, trade-offs, and a verification path instead of surface-level answers.
| Skill | Scope |
|---|---|
skills/software-architecture-engineering-uap | Design, evaluate, document, modernize, and refactor software architecture and code structure; microservices patterns; clean code; IAM/auth; NFR tactics; ADR/RFC; implementation planning. |
skills/uiux-frontend-engineering | Research, design, implement, and validate UI/UX and frontend interfaces for web, mobile, responsive apps, design systems, theme tokens, and XR/spatial surfaces. |
skills/data-architect-engineering | Design, review, migrate, optimize, secure, and operate SQL, NoSQL, vector, graph, search, cache, big-data, lakehouse, and warehouse systems. |
skills/device-driver-development | Linux kernel device drivers, modules, device tree bindings, bus clients, DMA/IRQ/memory paths, bring-up, and upstreaming. |
skills/realtime-systems-coding | Concurrent, parallel, distributed, real-time, and network-dataplane systems; locks/monitors, RMS/EDF scheduling, DPDK/XDP, kernel bypass. |
skills/tech-leadership | CTO and technology leadership: strategy, cybersecurity, requirements, framework selection, people, delivery, risk, KPIs, roadmap, budgeting, and governance. |
skills/academic-research-journal | Evaluate, design, draft, and revise academic articles, literature/systematic reviews, journal submissions, and source/integrity audits. |
The repository is one plugin. Skills live under skills/, and each platform reads its own manifest:
.claude-plugin/
plugin.json # Claude Code plugin metadata
marketplace.json # Claude Code marketplace entry
.codex-plugin/
plugin.json # Codex plugin metadata + interface
gemini-extension.json # Antigravity (Gemini) extension manifest
GEMINI.md # Antigravity context file (includes AGENTS.md)
AGENTS.md # Cross-platform agent orientation
hooks/
hooks.json # SessionStart orientation hook (Claude Code)
session-start.sh # emits the skillset orientation as additionalContext
.mcp.json # MCP servers (Context7, GitHub, Postgres, arXiv)
LICENSE # MIT
skills/
<skill>/
SKILL.md # Claude skill manifest (frontmatter: name, description)
agent.md # Agent identity and operating model
agents/openai.yaml # Codex interface config and default prompt
tasks/ # Step-by-step task workflows
references/ # Domain reference material
scripts/ # Helper scripts (static scans, probes)
commands/ # Slash commands (where applicable)claude plugin marketplace add oghie/skillsets
claude plugin install skillsetsSkills under skills/ auto-discover; Claude reads each SKILL.md and the description field drives auto-invocation.
Install the plugin from the same repository (oghie/skillsets). Codex reads .codex-plugin/plugin.json, which points at ./skills/ and exposes the bundled skills through its interface block.
Install this repository as a Gemini extension. Antigravity reads gemini-extension.json and loads GEMINI.md, which pulls in AGENTS.md for shared agent orientation.
The plugin ships a .mcp.json with four optional MCP servers. They are lazy-loaded and only matter for skills that benefit from live external data; each needs its own runtime and, where noted, configuration. A server that is not configured simply fails to start and shows up in /mcp — it does not block a session.
| Server | Backs | Runtime | Configuration |
|---|---|---|---|
context7 | All engineering skills (current library/API docs) | Node 18+ (npx) | None required; set CONTEXT7_API_KEY for higher rate limits. |
github | tech-leadership, software-architecture (repos, issues, PRs) | none (remote HTTP) | OAuth on first use against https://api.githubcopilot.com/mcp/. |
postgres | data-architect-engineering (schema, EXPLAIN, index tuning) | uv (uvx postgres-mcp) | Optional. Set DATABASE_URI to your connection string to activate it; if unset it stays inactive and the plugin loads normally. Runs in restricted (read-only) mode. |
arxiv | academic-research-journal (paper search, source audits) | uv (uvx arxiv-mcp-server) | None required. |
Notes:
postgres uses Postgres MCP Pro (postgres-mcp), not the archived@modelcontextprotocol/server-postgres. restricted access mode wraps queries in read-only transactions.
uv (https://docs.astral.sh/uv/) for the postgres and arxiv servers, andNode 18+ for context7.
.mcp.json.Each skill auto-surfaces by name and description. When a task matches a skill's description, the agent reads that skill's skills/<name>/SKILL.md and follows its workflow. All runtimes read the same tasks/, references/, and scripts/, so behavior stays consistent regardless of the host.
If you have none of these tools yet, pick a platform and follow two steps: install the CLI, then install skillsets.
Claude Code
curl -fsSL https://claude.ai/install.sh | bashirm https://claude.ai/install.ps1 | iexnpm install -g @anthropic-ai/claude-codeclaude --version claude plugin marketplace add oghie/skillsets
claude plugin install skillsetsCodex
npm install -g @openai/codexbrew install codexcodex --version codex plugin marketplace add oghie/skillsets
codex plugin install skillsetsOr run /plugins inside Codex to browse and install it.
Antigravity
https://antigravity.google, then sign in.
gemini extensions install https://github.com/oghie/skillsetsAntigravity loads the repo's gemini-extension.json and GEMINI.md. Update later with gemini extensions update skillsets.
After installing (all platforms)
the skill domains; the matching skill loads on its own.
context7, and uv forpostgres and arxiv (see MCP servers). Quick installs:
brew install node uvcurl -LsSf https://astral.sh/uv/install.sh | shirm https://astral.sh/uv/install.ps1 | iexrecommending anything.
current sources, not assumed.
references/ is the source of truth per domain; tasks/ orchestrates the workflow.~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.