agent-skills

Optimize Hermes memory files (MEMORY.md and USER.md) through signal/noise analysis and deep user interviews. Distinguish operational facts from fluff, and build meaningful user profiles.

End-to-end Electron desktop app expertise — plan, architect, secure, build, package, sign, notarize, auto-update, OS-integrate cross-platform (mac/Windows/Linux) apps incl. AI/LLM-powered. Use when user says "build electron app", "electron desktop app", "electron security", "contextBridge", "preload", "electron IPC", "electron packaging", "code sign electron", "notarize electron", "electron auto-update", "electron-vite", "electron-forge", "electron-builder", "electron fuses", "ASAR integrity", "safeStorage", "deep link", "custom protocol", "electron OAuth PKCE", "utilityProcess", "MessageChannelMain", "electron tray", "electron MCP", "AI desktop app", "Chromium upgrade", "Tauri vs Electron", "universal binary macOS", or shares an Electron project, package.json with electron, .dmg/.exe/.AppImage/.MSIX build issues. Do NOT use for pure web apps, browser extensions, or pure Tauri/Wails projects.

Use when authoring, refactoring, auditing, or learning to build any agent skill — for Claude Code, the Claude API, claude.ai, the Claude Agent SDK, Hermes Agent (Nous Research), OpenClaw, Codex, Cursor, Cline, or Antigravity. Use when the user says any of "make a skill", "write a skill", "skill for [tool]", "skill best practices", "fix this skill", "review my SKILL.md", "skill description not triggering", "skill discovery", "skill frontmatter", "progressive disclosure", "agent skill format", "claude skill", "SKILL.md template", "Hermes skill", "skillception", "meta skill". Use even when the user does not say the word "skill" but is clearly authoring a SKILL.md, packaging an onboarding guide for an agent, or trying to make an LLM consistently follow a multi-step procedure. Use proactively whenever a workflow has been repeated more than once. This skill is the canonical, self-contained authority on skill authoring; load it first, follow it exactly.

Use when user shares x.com/twitter.com URLs, asks to "read tweet", "search twitter", "check mentions", "timeline", "bird", or any tweet-related actions. Read tweets, search, and browse Twitter/X timelines via bird CLI. Do NOT use for general web browsing or non-Twitter sites.

Orchestrate OpenAI Codex CLI agents via tmux sessions for parallel coding tasks. Use when user says "spawn codex", "delegate to codex", "codex agent", "parallel coding", or for complex multi-step coding that benefits from agent swarm execution. Do NOT use for simple single-file edits or when user explicitly wants a different agent.

Persistent long-term research tracker that builds dated Markdown timelines for topics/people. Use when user says "track this over time", "research timeline", "last365", "persistent profile", "save research history", or requests multi-session tracking on a topic. Do NOT use for one-off quick searches without persistence.

Enforce NASA JPL "Power of 10" safety-critical coding rules on C/C++, Python, JS, TS, or Go code. Use when user says "apply nasa rules", "check power of 10", "nasa coding standards", "safety-critical audit", or "verify code reliability". Do NOT use for general code review without safety-critical context.