zm2231

PersonalOS skill: account-management

API endpoint testing, mocking, and integration testing for robust API development.

Optimize images, icons, and design assets for performance, quality, and developer handoff

Plan, track, and optimize budgets to maximize value and financial performance

CI/CD pipeline optimization, build failures, and deployment automation for DevOps workflows.

Code quality improvement, technical debt reduction, and refactoring patterns for maintainable software.

Comprehensive code review with quality checks, best practices, and actionable feedback

Lightweight customer relationship management for tracking contacts, deals, and interactions

Ingestion and indexing agent for the BTD Knowledge Engine. Use this skill when content needs to be added, indexed, or linked to user profiles. Handles the full pipeline: register creators, scan catalogs, ingest content, compile wiki pages, rebuild search index. Can be invoked as a subagent from another Claude Code instance.

Structured intake interview for new BTD Knowledge Engine users. Use this skill whenever someone wants to start a new project, learn a new domain, figure out what to build, or asks for help getting started with anything. Triggers on: "I want to build X", "help me learn Y", "I have an idea", "where should I start", "I'm stuck", "what should I do next", "I'm new here", or any expression of a goal without a clear plan. Also use when someone asks a question but has no profile yet (check with profile.js list first). Do NOT use for returning users who already have a profile — use btd-reentry instead.

Decide what content to ingest into the BTD Knowledge Engine. Use this skill when someone asks "what should we add," "who else should be in the corpus," "what's missing," "we need more on [topic]," "find creators for [domain]," or when evaluating whether specific content is worth ingesting. Also use after reviewing user profiles to identify corpus gaps; if intake interviews reveal topics the corpus can't serve, this skill figures out what to add. Triggers on any content strategy, corpus planning, or "what to ingest next" question.

Help a user build their personal knowledge corpus from scratch. Use this skill when someone says "I want to add my own sources," "set up my instance," "who should I follow," "I want to track [creator]," "add my favorite podcast," "I read this newsletter," "index my notes," or anything about populating their local corpus with content they care about. Also triggers when someone has an empty local/ directory. This skill defaults to --local (personal corpus).

Compile raw content into the wiki knowledge base. Use this skill whenever new content has been ingested, when someone asks to "update the wiki", "compile new sources", "what's in the knowledge base", or when compile-wiki.js shows pending sources. Also use after any batch ingestion, after adding a new creator, or when the wiki lint shows gaps. This is maintenance work — do it proactively, not just when asked.

Set up, publish to, and manage THIS Cloudflare Worker docs site (you are inside a docs-site-template checkout). Use when the user wants to do first-time setup of this cloned site, add/publish a doc (HTML, Markdown, PDF, or a pre-built SPA), set a per-document password for a client, or install the per-site "share" skill. Deploy target is Cloudflare, never a VPS. The index page is generated from git + _meta.json. To stand up a NEW site from scratch, use "new-docs-site".

Spin up a brand-new gated docs site on Cloudflare from the docs-site-template. Use when the user wants to stand up a new shared/gated/private documents site, a client deliverable portal, or says things like "spin up a new docs site", "create a gated docs site", "new shared docs site", "stand up a docs portal on Cloudflare", "make a password-protected docs site". This BUILDS a new site (clone the template, configure, deploy). Deploy target is Cloudflare, never a VPS. To add a doc to an ALREADY-built site, use that site's generated "share-<slug>" skill, or the in-repo "docs-site" skill.

Operate AgentHail across Claude, Codex, and Notion: discover sessions, create Notion threads, send and observe work, manage queues, aliases, channels, relay subscriptions, daemon notifications, and the dashboard.

Trace how one repository actually runs and maintain its local .agent/process-map.md as cited agent context. Use when a user asks how a repository works, when the map is missing or stale, or when an operator requests a map refresh.

Search LEANN indexes through CLI, MCP, or Python. Use for `leann search`, `ask`, `react`, metadata filters, hybrid BM25/vector search, temporal queries, sparse-filter diagnostics, result diversification, context windows, facets, and batched multi-search.

Build, update, rebuild, watch, or list LEANN indexes. Use for `leann build`, `build-jsonl`, `rebuild`, `watch`, source readers, backend choice, incremental behavior, metadata stamping, folder filters, embedding cache, and AST-aware code indexing.

Add, remove, or debug MCP servers in an mcporter-serve setup. Use when wiring a new MCP tool, removing one, checking status, or debugging why tools aren't showing. Triggers: - "add X as an MCP" - "register X in mcporter" - "wire X as a tool" - "remove X from mcporter" - "why isn't X showing as a tool" - "debug MCP" - "check MCP status"

Install, start, and configure mcporter-serve, the shared per-server MCP gateway. Use when setting it up for the first time, importing existing MCP servers from Claude/Codex/Cursor, starting or restarting the daemon and serve bridge, wiring per-server endpoints into a client, or debugging why an endpoint is down. Triggers: - "set up mcporter-serve" / "install the mcp gateway" - "import my mcp servers" - "start / restart the mcporter daemon" - "wire per-server mcp endpoints into claude" - "why isn't the serve endpoint responding"

Operate AgentHail across Claude, Codex, and Notion: discover sessions, create Codex and Notion threads, send and observe work, manage queues, aliases, channels, relay subscriptions, daemon notifications, and the dashboard.

Generate self-contained HTML pages that visually explain systems, code, plans, and data. Produces documents, briefs, memos, diagrams, audits, recaps, comparison tables, slide decks, multi-doc sites. Triggers - "diagram", "visualize", "render as a page", "build a brief", "memo for the team", "deck about", "audit", "recap", "fact-check", "plan review", "diff review", "small site I'll edit". Also fires proactively when about to render an ASCII table with 4+ rows or 3+ columns (generate HTML instead). NOT for applications, interactive UIs, landing pages, brand sites, dashboards-as-products, or component libraries. NOT for one-line answers, tiny markdown tables, code-as-artifact responses, or discussion-not-document conversations.

Interface design with craft and consistency — dashboards, apps, tools, landing pages, marketing sites. Any UI that needs a design system, memory, and enforcement.

Use when the user wants Codex to communicate with Claude Code sessions through claude-rc, send a message to an already-running Claude agent, ask Claude for output, fetch Claude's reply, list or resolve RC-enabled Claude sessions, check Remote Control health, or coordinate work between Codex and Claude without tmux or manual copy-paste.

Operate undertone, a local on-device audio transcript producer. Use when ingesting audio from local files, YouTube, podcasts, RSS feeds, direct media URLs, article/web media pages, Quill, or Google Meet; running transcription, diarization, or FluidAudio/pyannote engine selection; producing, re-enriching, browsing, searching, or exporting transcripts; labeling, relabeling, merging, discarding, restoring, destroying, exporting, or adopting speaker voice fingerprints; running doctor, stats, models, or webhook checks; installing, packaging, or upgrading undertone. Routes to detailed references under references/.

Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when user needs to navigate websites, interact with web pages, fill forms, take screenshots, or extract information.