arterm-sedov

Crawl and sanitize external websites for LLM ingestion. Use when needing to crawl comindware.ru or cmwlab.com, sanitize scraped content, or manage the crawl→sanitize pipeline. Supports --fresh (from scratch) and --resume (checkpoint) workflows.

Use when you've completed a non-trivial task and need to document discoveries, patterns, or gotchas for future sessions. Also use to review the discovery log before starting similar work.

Transcribe meeting recordings with Google Gemini for future documentation work. Use when the user provides a video file and needs a structured Markdown transcript with timestamps, visual notes, and summary in .scratch/.

Use when working on the PHPKB cloning and post-clone migration workflow in this repository: creating a new PHPKB section for a new product version, publishing a new MkDocs article by cloning an adjacent PHPKB article, syncing changed for_kb_import_ru HTML back to PHPKB by kb-id (git diff batch), cloning PHPKB categories or articles, updating cloned PHPKB article/category links, migrating local docs IDs with clone mappings, fixing related topics after cloning, or analyzing/updating scripts in utilities/phpkb_cloning.

Use when writing or answering about N3/Turtle/RDF/Notation3/Triples in Comindware Platform

Reference files and patterns for writing C# scripts in Comindware Platform

Use when regenerating or ingesting the Comindware Platform LLM knowledge bundle in this repository: refreshing phpkb_content_rag from PHPKB, running phpkb_import_for_rag.py, running phpkb_ingest.py, or updating the LLM ingestion bundle.

Use when editing an existing KB article (docs/ru/**/*.md with kbId), rebuilding PHPKB HTML export, publishing to PHPKB, and committing. Covers the full edit→rebuild→publish→commit cycle. Triggers on "edit article", "update article", "republish", "rebuild and publish", "sync to PHPKB", or any task modifying an existing docs/ru Markdown file that has a kbId.

Use when working with Comindware Platform — connecting to platform, listing applications, exploring templates, managing records, querying data, creating or editing attributes, or any task requiring autonomous platform interaction. Triggers on platform operations, CMW queries, tenant management, rental lot operations, debt tracking, or record CRUD. Also triggers when user mentions working directly with API credentials, using manual HTTP approaches, or trying to bypass the tool layer. ALSO triggers when user mentions browser automation, UI-only features, visual verification, or accessing platform features not available via API.

Access Hugging Face Spaces logs via the HF API (SSE streaming). Use when the user needs to view, tail, or stream container logs or build logs from a Hugging Face Space (including cmw-copilot), debug Space deployments, inspect runtime errors, check build progress, inspect or update Space variables/secrets, restart a Space, or verify environment-driven integrations such as MCP. Triggers on mentions of "HF logs", "Space logs", "Hugging Face logs", "cmw-copilot logs", "container logs", "build logs", "Space debug", "HF variables", "Space variables", "Space secrets", or "HF_TOKEN".

Moved to instance repository. US FM hierarchy seed workflows are tenant-specific; use the copy under {instance_progress_dir}.

Create or update platform login accounts via System Core AccountService — Create, Edit, SetAccountPassword, AccountGroupService IncludeMembers, optional solution role. Triggers on seed users, administration accounts, AccountService, SetAccountPassword, SetUsername, account rename, Mbox, FullName, UAT login bootstrap. Agnostic for any CMW instance (CMW_BASE_URL). Does not cover batch migration workflows (those live in the project repo).

Fills or updates running process instance records (work orders, events, form.oa) on Comindware Platform via Records API or create_edit_record. Use for process instance, work order, running process, demo fill, or placeholder message text on in-progress rows. Instance-agnostic via CMW_BASE_URL. Prefer OpenAPI and GET/PUT before browser; browser only for workflow buttons/transitions.

Launch an existing Comindware Platform configuration backup via Web API (preferred) or admin UI fallback. Use when the user or migration plan requires a post-change backup, configuration backup, FR rollback snapshot, or after-batch safety backup between major themed migration edits. Agnostic for any instance via CMW_BASE_URL. Does not create backup configurations unless the user explicitly asks.

Switch Comindware Platform target instance (host/tenant) via CMW_BASE_URL and dotenv. Use when the user asks to switch instance, change host, connect to a different server, update CMW_BASE_URL, switch source vs target host, edit .env for another tenant, or verify credentials before bulk platform work. Agnostic for any instance pair. Does not duplicate general platform CRUD — see cmw-platform skill.

Links an existing platform login to a Staff (Sotrudniki) account-template employee row on Comindware Platform. Use when attaching accounts, Staff, Sotrudniki, account template records, employee login link, or UAT persona attach after account bootstrap. Instance-agnostic via CMW_BASE_URL. Prefer OpenAPI and agent tools before browser; UI Attach account modal is last resort when IncludeInContainer fails on a host.

Use for web search and content extraction when you want a free, unlimited, local solution. Priority: native websearch (simple) → searxng (comprehensive, free) → tavily (advanced features). Use when: user wants comprehensive results without API keys, mentions "free", "unlimited", "local", "SearXNG", "no API key", or searxng-docker-tavily-adapter stack is running. Falls back to tavily if advanced features needed (research synthesis, JS rendering).

Use for URL content extraction when you want a free, unlimited, local solution. Priority: searxng-extract (free, basic) → tavily-extract (advanced features). Use when: user wants page content without API keys, mentions "free", "local", "SearXNG", or searxng-docker-tavily-adapter stack is running. Falls back to tavily-extract if advanced features needed (JS rendering, query-focused chunking).

Use for web search when you need more than simple lookups AND want a free, unlimited solution. Priority: native websearch (simple) → searxng-search (comprehensive, free) → tavily-search (advanced features). Use when: user wants comprehensive results, mentions "free", "unlimited", "local", "no API key", "SearXNG", or searxng-docker-tavily-adapter stack is running. Falls back to tavily-search if advanced features needed (topic filtering, time ranges).

CMW Platform v5.0 knowledge: answer from authoritative documentation, not guesses. Use for customers, partners, and employees whenever the user asks about product behavior, administration, configuration, integrations, APIs, troubleshooting, or "what the documentation says"—even if they do not say "knowledge base" or "KB." Source order: (1) local Markdown under docs/ru only when this Git workspace's remotes match the canonical cbap-mkdocs-ru repo (see references/git_remote_allowlist.md); (2) CMW KB MCP tools if configured; (3) official kb.comindware.ru LLM ingestion .md; (4) GitHub raw on branch platform_v5 under docs/ru via raw.githubusercontent.com (bootstrap from toc.md and guides_toc.md, never the GitHub tree UI alone). Respond in the user's language; keep citations aligned with the source text. If nothing authoritative answers the question, say so—do not invent platform facts.

Comprehensive human-like web search and scraping meta-skill. Mimics human behavior to bypass blocks/rate-limits. Auto-installs dependencies, auto-repairs failures, cascades through 5 tiers: (1) Browser CLI (agent-browser/playwright), (2) Scraping APIs (crawl4ai/jina.ai), (3) Python scraper (unlimited), (4) Quick extract (searxng/tavily), (5) Discovery search. Includes Yandex search. For Telegram login: telegram-scraper. Use for any "search", "scrape", "extract", "find articles" request.

Use this skill whenever the user wants to fetch, sync, update, refresh, or index the Comindware MkDocs/PHPKB RAG corpora. This skill is specifically for the managed sparse clone workflow that keeps both V5 and V6 corpora under .reference-repos/cbap-mkdocs-ru and indexes them with rag_engine/scripts/build_index.py.

Conducts enterprise-grade research with multi-source synthesis, citation tracking, and verification. Produces citation-backed reports through a structured pipeline. Uses FREE providers by default (WebSearch, SearXNG). Triggers on "deep research", "comprehensive analysis", "research report", "compare X vs Y", "analyze trends", or "state of the art". Not for simple lookups, debugging, or questions answerable with 1-2 searches.

Choose between agent-browser and Playwright CLI for browser automation tasks.

Provides deterministic primitives for restructuring markdown documents with 100% reliability and coverage verification.

Fix Tavily MCP "Client Not Registered" OAuth authentication errors. Use this skill when encountering OAuth failures, invalid client_id errors, or Tavily MCP tools failing to authenticate. Clears cached credentials and triggers fresh OAuth authentication flow.

Extract TEXT content from Telegram channels (public, private, paywalled). Use this skill when the user wants to scrape Telegram channels, extract chat history, download messages from Telegram, or says "scrape Telegram", "get Telegram content", "extract from t.me". Focus on text extraction - OCR/vision only for browser automation (clicking, searching).